Overhaul memory allocation model

This commit makes several fundamental changes to memory handling:

- the frame allocator is now only an allocator for free frames, and does
  not track used frames.
- the frame allocator now stores its free list inside the free frames
  themselves, as a hybrid stack/span model.
  - This has the implication that all frames must currently fit within
    the offset area.
- kutil has a new allocator interface, which is the only allowed way for
  any code outside of src/kernel to allocate. Code under src/kernel
  _may_ use new/delete, but should prefer the allocator interface.
- the heap manager has become heap_allocator, which is merely an
  implementation of kutil::allocator which doles out sections of a given
  address range.
- the heap manager now only writes block headers when necessary,
  avoiding page faults until they're actually needed
- page_manager now has a page fault handler, which checks with the
  address_manager to see if the address is known, and provides a frame
  mapping if it is, allowing heap manager to work with its entire
  address size from the start. (Currently 32GiB.)

.gitignore

@@ -1,7 +1,8 @@
 .lock*
-build
+/build*
 *.bak
 tags
-.gdbinit
 popcorn.log
-.waf-*
+*.o
+*.a
+sysroot

LICENSE.md

@@ -2,7 +2,8 @@
 
 Popcorn itself is released under the terms of the MIT license:
 
-> Copyright © 2018 Justin C. Miller, http://devjustinian.com <justin@devjustinian.com>
+> Copyright © 2018 Justin C. Miller, https://devjustinian.com
+> <justin@devjustinian.com>
 > 
 > Permission is hereby granted, free of charge, to any person obtaining a copy
 > of this software and associated documentation files (the “Software”), to deal
@@ -19,8 +20,8 @@ Popcorn itself is released under the terms of the MIT license:
 > FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 > AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 > LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-> OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
-> THE SOFTWARE.
+> OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+> SOFTWARE.
 
 # Included works
 
@@ -28,7 +29,7 @@ Popcorn includes and/or is derived from a number of other works, listed here.
 
 ## Catch2
 
-Popcorn uses [Catch2](http://github.com/catchorg/Catch2) for testing. Catch2 is
+Popcorn uses [Catch2](https://github.com/catchorg/Catch2) for testing. Catch2 is
 released under the terms of the Boost Software license:
 
 > Boost Software License - Version 1.0 - August 17th, 2003
@@ -55,44 +56,57 @@ released under the terms of the Boost Software license:
 > ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 > DEALINGS IN THE SOFTWARE.
 
-## Intel EFI Application Toolkit
+## cpptoml
 
-Popcorn's UEFI loader uses code from Intel's EFI Application toolkit. Relevant
-code includes license statements at the top of each file.
+Popcorn uses the [cpptoml](https://github.com/skystrife/cpptoml) library for
+parsing TOML configuration files. cpptoml is released under the terms of the
+MIT license:
 
-## Waf
+> Copyright (c) 2014 Chase Geigle
+>
+> Permission is hereby granted, free of charge, to any person obtaining a copy of
+> this software and associated documentation files (the "Software"), to deal in
+> the Software without restriction, including without limitation the rights to
+> use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
+> the Software, and to permit persons to whom the Software is furnished to do so,
+> subject to the following conditions:
+> 
+> The above copyright notice and this permission notice shall be included in all
+> copies or substantial portions of the Software.
+> 
+> THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+> IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
+> FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
+> COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
+> IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+> CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
-Popcorn's build system uses [Waf](https://waf.io/), which claims to be released
-under the BSD license. I could not find its specific license file, so I am
-reproducing a generic 3-clause BSD license (the most restrictive, so as not to
-assume any extra rights that may not actually be granted) for it here:
+## printf
 
-> Copyright © 2005-2018 Thomas Nagy
-> 
-> Redistribution and use in source and binary forms, with or without
-> modification, are permitted provided that the following conditions are met:
-> 
-> 1. Redistributions of source code must retain the above copyright notice, this
->    list of conditions and the following disclaimer.
-> 
-> 2. Redistributions in binary form must reproduce the above copyright notice,
->    this list of conditions and the following disclaimer in the documentation
->    and/or other materials provided with the distribution.
-> 
-> 3. Neither the name of the copyright holder nor the names of its contributors
->    may be used to endorse or promote products derived from this software
->    without specific prior written permission.
-> 
-> THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
-> ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-> WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
-> DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
-> FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
-> DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
-> SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
-> CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
-> OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-> OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+Popcorn uses 's tiny [printf](https://github.com/mpaland/printf) library for its
+`*printf()` functions, which is also released under the terms of the MIT license:
+
+> The MIT License (MIT)
+>
+> Copyright (c) 2014 Marco Paland
+>
+> Permission is hereby granted, free of charge, to any person obtaining a copy
+> of this software and associated documentation files (the "Software"), to deal
+> in the Software without restriction, including without limitation the rights
+> to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+> copies of the Software, and to permit persons to whom the Software is
+> furnished to do so, subject to the following conditions:
+>
+> The above copyright notice and this permission notice shall be included in all
+> copies or substantial portions of the Software.
+>
+> THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+> IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+> FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+> AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+> LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+> OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+> SOFTWARE.
 
 ## GNU-EFI
 
@@ -130,3 +144,7 @@ to assume any extra rights that may not actually be granted) for it here:
 > OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 > OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
+## Intel EFI Application Toolkit
+
+Popcorn's UEFI loader uses code from Intel's EFI Application toolkit. Relevant
+code includes license statements at the top of each file.

NOTES.md

@@ -2,23 +2,37 @@
 
 ## TODO
 
-- Better page-allocation model
+- Paging manager
+  - Copy-on-write pages
+  - Better page-allocation model?
 - Allow for more than one IOAPIC in ACPI module
   - The objects get created, but GSI lookup only uses the one at index 0
-- Slab allocator for kernel structures
 - mark kernel memory pages global
-- lock `memory_manager` and `page_manager` structures
 - Serial out based on circular/bip biffer and interrupts, not spinning on
   `write_ready()`
 - Split out more code into kutil for testing
-
-
+- AHCI / MSI interrupts on Vbox break?
+- FXSAVE to save XMM registers.
+  - optimization using #NM (0x7) to detect SSE usage
+- Clean up of process memory maps
+- Better stack tracer
+- Bootloader rewrite
+  - C++ and sharing library code for ELF, initrd, etc
+  - Parse initrd and pre-load certain ELF images, eg the process loader process?
+  - Do initial memory bootstrap?
+- Calling global ctors
 - Device Tree
   - Actual serial driver
   - Disk driver
     - File system
-- Memory map swapping
-  - Multiprocessing
-    - Processes in Ring 3
-- Stack tracer
-- build system upgrade (CMake / Waf / etc)
+- Multiprocessing
+  - Fast syscalls using syscall/sysret
+
+### Build
+
+- Clean up build generator and its templates
+  - More robust objects to represent modules & targets to pass to templates
+  - Read project setup from a simple JSON/TOML/etc
+  - Better compartmentalizing when doing template inheritance
+- Move to LLD as sysroot linker
+

README.md

@@ -27,14 +27,45 @@ The design goals of the project are:
 
 ## Building
 
-Popcorn uses the `waf` build tool, which is included in the repo. The other
-requirements are:
+Popcorn uses the [Ninja][] build tool, and generates the build files for it
+with a custom tool called [Bonnibel][]. Bonnibel requires [Python 3][] and can
+be downloaded with `pip`:
 
-* python (to run waf)
+```
+pip3 install bonnibel
+```
+
+[Ninja]:    https://ninja-build.org
+[Bonnibel]: https://github.com/justinian/bonnibel
+[Python 3]: https://python.org
+
+Requrirements:
+
+* python 3 (for installing and running Bonnibel)
 * clang
-* nasm
+* mtools
+* ninja
+* curl for downloading the toolchain
+
+### Setting up the cross toolchain
+
+If you have `clang` and `curl` installed, runing the `scripts/build_sysroot_clang.sh`
+script will download and build a nasm/binutils/LLVM toolchain configured for building
+Popcorn host binaries.
+
+### Building and running Popcorn
+
+Once the toolchain has been set up, running Bonnibel's `pb` command will set up the
+build configuration, and `ninja -C build` will actually run the build.  If you
+have `qemu-system-x86_64` installed, the `qemu.sh` script will to run Popcorn
+in QEMU `-nographic` mode.
+
+I personally run this either from a real debian amd64 testing/buster machine or
+a windows WSL debian testing/buster installation. The following should be
+enough to set up such a system to build the kernel:
+
+    sudo apt install qemu-system-x86 nasm clang-6.0 mtools python3-pip curl
+    sudo update-alternatives /usr/bin/clang clang /usr/bin/clang-6.0 1000
+    sudo update-alternatives /usr/bin/clang++ clang++ /usr/bin/clang++-6.0 1000
+    sudo pip3 install bonnibel
 
-After cloning, run `waf configure`. Then you can run `waf build` to build the
-project, and `waf test` to run the tests. A floppy disk image will be built in
-`build/popcorn.img`. If you have `qemu-system-x86_64` installed, then you can
-run `waf qemu` to run it in `-nographic` mode.

assets/debugging/popcorn.elf-gdb.py

@@ -0,0 +1,31 @@
+import gdb
+
+class PrintStackCommand(gdb.Command):
+    def __init__(self):
+        super().__init__("popc_stack", gdb.COMMAND_DATA)
+
+    def invoke(self, arg, from_tty):
+        args = gdb.string_to_argv(arg)
+
+        base = "$rsp"
+        if len(args) > 0:
+            base = args[0]
+
+        depth = 22
+        if len(args) > 1:
+            depth = int(args[1])
+
+        for i in range(depth-1, -1, -1):
+            offset = i * 8
+            base_addr = gdb.parse_and_eval(base)
+            value = gdb.parse_and_eval(f"*(uint64_t*)({base} + {offset})")
+            print("{:016x} (+{:04x}): {:016x}".format(int(base_addr) + offset, offset, int(value)))
+
+
+PrintStackCommand()
+
+import time
+time.sleep(3.5)
+gdb.execute("target remote :1234")
+gdb.execute("set waiting = false")
+gdb.execute("display/i $rip")

assets/initrd.toml

@@ -0,0 +1,23 @@
+# This is the manifest for the initial ramdisk, read by the `makerd` tool.
+# The contents should be a table array of files to add to the ramdistk:
+#
+#  [[files]]
+#  dest = "foo.bar"             # Name of the file in the ramdisk
+#  source = "build/foo/foo.bar" # Location of the file from the project root
+#  executable = true            # Optional, default false. Whether this is an
+#                               # initial application for the kernel to execute
+#                               # on startup
+
+[[files]]
+dest = "screenfont.psf"
+source = "../assets/fonts/tamsyn8x16r.psf"
+
+[[files]]
+dest = "nulldrv1"
+source = "user/nulldrv"
+executable = true
+
+[[files]]
+dest = "nulldrv2"
+source = "user/nulldrv"
+executable = true

modules.yaml

@@ -0,0 +1,132 @@
+name: Popcorn
+templates: scripts/templates
+modules:
+    kernel:
+        output: popcorn.elf
+        target: host
+        deps:
+            - elf
+            - initrd
+            - kutil
+        includes:
+            - src/kernel
+        source:
+            - src/kernel/crti.s
+            - src/kernel/apic.cpp
+            - src/kernel/assert.cpp
+            - src/kernel/boot.s
+            - src/kernel/console.cpp
+            - src/kernel/cpprt.cpp
+            - src/kernel/cpu.cpp
+            - src/kernel/debug.cpp
+            - src/kernel/debug.s
+            - src/kernel/device_manager.cpp
+            - src/kernel/font.cpp
+            - src/kernel/frame_allocator.cpp
+            - src/kernel/fs/gpt.cpp
+            - src/kernel/gdt.cpp
+            - src/kernel/gdt.s
+            - src/kernel/interrupts.cpp
+            - src/kernel/interrupts.s
+            - src/kernel/io.cpp
+            - src/kernel/loader.s
+            - src/kernel/log.cpp
+            - src/kernel/main.cpp
+            - src/kernel/memory_bootstrap.cpp
+            - src/kernel/msr.cpp
+            - src/kernel/page_manager.cpp
+            - src/kernel/pci.cpp
+            - src/kernel/process.cpp
+            - src/kernel/scheduler.cpp
+            - src/kernel/screen.cpp
+            - src/kernel/serial.cpp
+            - src/kernel/syscall.cpp
+            - src/kernel/syscall.s
+            - src/kernel/syscalls/exit.cpp
+            - src/kernel/syscalls/fork.cpp
+            - src/kernel/syscalls/getpid.cpp
+            - src/kernel/syscalls/message.cpp
+            - src/kernel/syscalls/noop.cpp
+            - src/kernel/syscalls/pause.cpp
+            - src/kernel/syscalls/sleep.cpp
+            - src/kernel/task.s
+            - src/kernel/crtn.s
+
+    boot:
+        kind: exe
+        target: boot
+        output: boot.elf
+        source:
+            - src/boot/crt0.s
+            - src/boot/console.cpp
+            - src/boot/guids.cpp
+            - src/boot/loader.cpp
+            - src/boot/main.cpp
+            - src/boot/memory.cpp
+            - src/boot/reloc.cpp
+            - src/boot/utility.cpp
+
+    nulldrv:
+        kind: exe
+        target: user
+        output: nulldrv
+        source:
+            - src/drivers/nulldrv/main.cpp
+            - src/drivers/nulldrv/main.s
+
+    elf:
+        kind: lib
+        output: libelf.a
+        deps:
+            - kutil
+        includes:
+            - src/libraries/elf/include
+        source:
+            - src/libraries/elf/elf.cpp
+
+    initrd:
+        kind: lib
+        output: libinitrd.a
+        deps:
+            - kutil
+        includes:
+            - src/libraries/initrd/include
+        source:
+            - src/libraries/initrd/initrd.cpp
+
+    kutil:
+        kind: lib
+        output: libkutil.a
+        includes:
+            - src/libraries/kutil/include
+        source:
+            - src/libraries/kutil/assert.cpp
+            - src/libraries/kutil/bip_buffer.cpp
+            - src/libraries/kutil/heap_allocator.cpp
+            - src/libraries/kutil/logger.cpp
+            - src/libraries/kutil/memory.cpp
+            - src/libraries/kutil/printf.c
+
+    makerd:
+        kind: exe
+        target: native
+        output: makerd
+        deps:
+            - initrd
+        source:
+            - src/tools/makerd/entry.cpp
+            - src/tools/makerd/main.cpp
+
+    tests:
+        kind: exe
+        target: native
+        output: tests
+        deps:
+            - kutil
+        source:
+            - src/tests/address_manager.cpp
+            - src/tests/constexpr_hash.cpp
+            - src/tests/linked_list.cpp
+            - src/tests/logger.cpp
+            - src/tests/heap_allocator.cpp
+            - src/tests/main.cpp

qemu.sh

@@ -0,0 +1,56 @@
+#!/usr/bin/env bash
+
+build="$(dirname $0)/build"
+assets="$(dirname $0)/assets"
+debug=""
+flash_name="ovmf_vars"
+gfx="-nographic"
+kvm=""
+
+for arg in $@; do
+	case "${arg}" in
+		--debug)
+			debug="-s"
+			flash_name="ovmf_vars_d"
+			;;
+		--gfx)
+			gfx=""
+			;;
+		--kvm)
+			kvm="-enable-kvm"
+			;;
+		*)
+			build="${arg}"
+			;;
+	esac
+done
+
+if [[ ! -c /dev/kvm ]]; then
+	kvm=""
+fi
+
+if ! ninja -C "${build}"; then
+	exit 1
+fi
+
+if [[ -n $TMUX ]]; then
+	if [[ -n $debug ]]; then
+		tmux split-window "gdb ${build}/popcorn.elf" &
+	else
+		tmux split-window -l 10 "sleep 1; telnet localhost 45454" &
+	fi
+fi
+
+exec qemu-system-x86_64 \
+	-drive "if=pflash,format=raw,readonly,file=${assets}/ovmf/x64/ovmf_code.fd" \
+	-drive "if=pflash,format=raw,file=${build}/${flash_name}.fd" \
+	-drive "format=raw,file=${build}/popcorn.img" \
+	-monitor telnet:localhost:45454,server,nowait \
+	-smp 4 \
+	-m 512 \
+	-d mmu,int,guest_errors \
+	-D popcorn.log \
+	-cpu Broadwell \
+	-M q35 \
+	-no-reboot \
+	$gfx $kvm $debug

scripts/build_sysroot_clang.sh

@@ -0,0 +1,221 @@
+#!/usr/bin/env bash
+
+TARGET="x86_64-elf"
+NASM_VERSION="2.13.03"
+BINUTILS_VERSION="2.31.1"
+LLVM_BRANCH="release_80"
+
+TOOLS="clang" # lld libunwind libcxxabi libcxx"
+PROJECTS="compiler-rt libcxxabi libcxx libunwind"
+#RUNTIMES="compiler-rt libcxxabi libcxx libunwind"
+
+set -e
+
+SYSROOT=$(realpath "$(dirname $0)/../sysroot")
+WORK=$(realpath "$(dirname $0)/sysroot")
+mkdir -p "${SYSROOT}"
+mkdir -p "${WORK}"
+
+export CC=clang
+export CXX=clang++
+
+function build_nasm() {
+	if [[ ! -d "${WORK}/nasm-${NASM_VERSION}" ]]; then
+		echo "Downloading NASM..."
+		tarball="nasm-${NASM_VERSION}.tar.gz"
+		curl -sSOL "https://www.nasm.us/pub/nasm/releasebuilds/${NASM_VERSION}/${tarball}"
+		tar xzf "${tarball}" -C "${WORK}" && rm "${tarball}"
+	fi
+
+	mkdir -p "${WORK}/build/nasm"
+	pushd "${WORK}/build/nasm"
+
+	if [[ ! -f "${WORK}/build/nasm/config.cache" ]]; then
+		echo "Configuring NASM..."
+		"${WORK}/nasm-${NASM_VERSION}/configure" \
+			--quiet \
+			--config-cache \
+			--disable-werror \
+			--prefix="${SYSROOT}" \
+			--srcdir="${WORK}/nasm-${NASM_VERSION}"
+	fi
+
+	echo "Building NASM..."
+	(make -j && make install) > "${WORK}/build/nasm_build.log"
+	popd
+}
+
+function build_binutils() {
+	if [[ ! -d "${WORK}/binutils-${BINUTILS_VERSION}" ]]; then
+		echo "Downloading binutils..."
+		tarball="binutils-${BINUTILS_VERSION}.tar.gz"
+		curl -sSOL "https://ftp.gnu.org/gnu/binutils/${tarball}"
+		tar xzf "${tarball}" -C "${WORK}" && rm "${tarball}"
+	fi
+
+	mkdir -p "${WORK}/build/binutils"
+	pushd "${WORK}/build/binutils"
+
+	if [[ ! -f "${WORK}/build/binutils/config.cache" ]]; then
+		echo "Configuring binutils..."
+		"${WORK}/binutils-${BINUTILS_VERSION}/configure" \
+			--quiet \
+			--config-cache \
+			--target="${TARGET}" \
+			--prefix="${SYSROOT}" \
+			--with-sysroot="${SYSROOT}" \
+			--with-lib-path="${SYSROOT}/lib" \
+			--disable-nls \
+			--disable-werror
+	fi
+
+	echo "Building binutils..."
+	(make -j && make install) > "${WORK}/build/binutils_build.log"
+	popd
+}
+
+function build_llvm() {
+	if [[ ! -d "${WORK}/llvm" ]]; then
+		echo "Downloading LLVM..."
+		git clone -q \
+			--branch "${LLVM_BRANCH}" \
+			--depth 1 \
+			"https://git.llvm.org/git/llvm.git" "${WORK}/llvm"
+	fi
+
+	for tool in ${TOOLS}; do
+		if [[ ! -d "${WORK}/llvm/tools/${tool}" ]]; then
+			echo "Downloading ${tool}..."
+			git clone -q \
+				--branch "${LLVM_BRANCH}" \
+				--depth 1 \
+				"https://git.llvm.org/git/${tool}.git" "${WORK}/llvm/tools/${tool}"
+		fi
+	done
+
+	if [[ ! -d "${WORK}/llvm/tools/clang/tools/extra" ]]; then
+		echo "Downloading clang-tools-extra..."
+		git clone -q \
+			--branch "${LLVM_BRANCH}" \
+			--depth 1 \
+			"https://git.llvm.org/git/clang-tools-extra.git" "${WORK}/llvm/tools/clang/tools/extra"
+	fi
+
+	for proj in ${PROJECTS}; do
+		if [[ ! -d "${WORK}/llvm/projects/${proj}" ]]; then
+			echo "Downloading ${proj}..."
+			git clone -q \
+				--branch "${LLVM_BRANCH}" \
+				--depth 1 \
+				"https://git.llvm.org/git/${proj}.git" "${WORK}/llvm/projects/${proj}"
+		fi
+	done
+
+	for proj in ${RUNTIMES}; do
+		if [[ ! -d "${WORK}/llvm/runtimes/${proj}" ]]; then
+			echo "Downloading ${proj}..."
+			git clone -q \
+				--branch "${LLVM_BRANCH}" \
+				--depth 1 \
+				"https://git.llvm.org/git/${proj}.git" "${WORK}/llvm/runtime/${proj}"
+		fi
+	done
+
+	mkdir -p "${WORK}/build/llvm"
+	pushd "${WORK}/build/llvm"
+
+	echo "Configuring LLVM..."
+
+	cmake -G Ninja \
+		-DCLANG_DEFAULT_RTLIB=compiler-rt \
+		-DCLANG_DEFAULT_STD_C=c11 \
+		-DCLANG_DEFAULT_STD_CXX=cxx14 \
+		-DCMAKE_BUILD_TYPE=Release \
+		-DCMAKE_C_COMPILER="clang" \
+		-DCMAKE_CXX_COMPILER="clang++" \
+		-DCMAKE_CXX_FLAGS="-Wno-unused-parameter -D_LIBCPP_HAS_NO_ALIGNED_ALLOCATION -D_LIBUNWIND_IS_BAREMETAL=1 -U_LIBUNWIND_SUPPORT_DWARF_UNWIND" \
+		-DCMAKE_INSTALL_PREFIX="${SYSROOT}" \
+		-DCMAKE_MAKE_PROGRAM=`which ninja` \
+		-DDEFAULT_SYSROOT="${SYSROOT}" \
+		-DLIBCXX_CXX_ABI=libcxxabi \
+		-DLIBCXX_CXX_ABI_INCLUDE_PATHS="${WORK}/llvm/projects/libcxxabi/include" \
+		-DLIBCXX_CXX_ABI_LIBRARY_PATH=lib \
+		-DLIBCXX_ENABLE_EXPERIMENTAL_LIBRARY=OFF \
+		-DLIBCXX_ENABLE_NEW_DELETE_DEFINITIONS=ON \
+		-DLIBCXX_ENABLE_SHARED=OFF \
+		-DLIBCXX_ENABLE_STATIC_ABI_LIBRARY=ON \
+		-DLIBCXX_ENABLE_THREADS=OFF \
+		-DLIBCXX_INCLUDE_BENCHMARKS=OFF \
+		-DLIBCXX_USE_COMPILER_RT=ON \
+		-DLIBCXXABI_ENABLE_NEW_DELETE_DEFINITIONS=OFF \
+		-DLIBCXXABI_ENABLE_SHARED=OFF \
+		-DLIBCXXABI_ENABLE_STATIC_UNWINDER=ON \
+		-DLIBCXXABI_ENABLE_THREADS=OFF \
+		-DLIBCXXABI_LIBCXX_PATH="${WORK}/llvm/projects/libcxx" \
+		-DLIBCXXABI_USE_COMPILER_RT=ON \
+		-DLIBCXXABI_USE_LLVM_UNWINDER=ON \
+		-DLIBUNWIND_ENABLE_SHARED=OFF \
+		-DLIBUNWIND_ENABLE_THREADS=OFF \
+		-DLIBUNWIND_USE_COMPILER_RT=ON \
+		-DLLVM_CONFIG_PATH="${SYSROOT}/bin/llvm-config" \
+		-DLLVM_DEFAULT_TARGET_TRIPLE="x86_64-unknown-elf" \
+		-DLLVM_ENABLE_LIBCXX=ON \
+		-DLLVM_ENABLE_PIC=OFF \
+		-DLLVM_ENABLE_THREADS=OFF \
+		-DLLVM_INSTALL_BINUTILS_SYMLINKS=ON \
+		-DLLVM_TARGETS_TO_BUILD="X86" \
+		${WORK}/llvm > cmake_configure.log
+
+	#  -DCMAKE_ASM_COMPILER=nasm \
+	#  -DCMAKE_LINKER="${SYSROOT}/bin/ld.lld" \
+	#  -DCOMPILER_RT_ENABLE_LLD=ON \
+	#  -DLIBCXX_ENABLE_LLD=ON \
+	#  -DLIBCXX_ENABLE_STATIC_UNWINDER=ON \
+	#  -DLIBCXXABI_ENABLE_LLD=ON \
+	#  -DLIBUNWIND_ENABLE_LLD=ON \
+	#  -DLLVM_ENABLE_LLD=ON \
+	#  -DLLVM_ENABLE_PROJECTS="libcxx;libcxxabi;libunwind;compiler-rt" \
+	#  -DCOMPILER_RT_BAREMETAL_BUILD=ON \
+	#  -DLIBCXXABI_BAREMETAL=ON \
+
+	echo "Building LLVM..."
+	ninja && ninja install
+	ninja cxx cxxabi compiler-rt
+	ninja install-compiler-rt install-cxx install-cxxabi
+	popd
+}
+
+function build_libc() {
+	if [[ ! -d "${WORK}/poplibc" ]]; then
+		echo "Downloading poplibc..."
+		git clone \
+			"https://github.com/justinian/poplibc.git" \
+			"${WORK}/poplibc"
+	else
+		echo "Updating poplibc..."
+		git -C "${WORK}/poplibc" pull
+	fi
+
+	pushd "${WORK}/poplibc"
+	echo "Building poplibc..."
+	make install PREFIX="${SYSROOT}"
+	popd
+}
+
+function update_links() {
+	for exe in `ls "${SYSROOT}/bin/${TARGET}-"*`; do
+		base=$(echo "$exe" | sed -e "s/${TARGET}-//")
+		ln -fs "${exe}" "${base}"
+	done
+}
+
+build_nasm
+build_binutils
+build_libc
+build_llvm
+update_links
+
+export CC="${SYSROOT}/bin/clang"
+export CXX="${SYSROOT}/bin/clang++"
+export LD="${SYSROOT}/bin/ld"
+build_libc

scripts/build_sysroot_gcc.sh

@@ -0,0 +1,186 @@
+#!/usr/bin/env bash
+
+TARGET="x86_64-elf"
+NASM_VERSION="2.14.02"
+GCC_VERSION="7.4.0"
+BINUTILS_VERSION="2.31.1"
+
+SYSROOT=$(realpath "$(dirname $0)/../sysroot")
+WORK=$(realpath "$(dirname $0)/sysroot")
+
+echo "Not currently supported"
+exit 1
+
+set -e
+mkdir -p "${SYSROOT}"
+mkdir -p "${WORK}"
+
+
+function build_nasm() {
+	if [[ ! -d "${WORK}/nasm-${NASM_VERSION}" ]]; then
+		echo "Downloading NASM..."
+		tarball="nasm-${NASM_VERSION}.tar.gz"
+		curl -sSOL "https://www.nasm.us/pub/nasm/releasebuilds/${NASM_VERSION}/${tarball}"
+		tar xzf "${tarball}" -C "${WORK}" && rm "${tarball}"
+	fi
+
+	mkdir -p "${WORK}/build/nasm"
+	pushd "${WORK}/build/nasm"
+
+	if [[ ! -f "${WORK}/build/nasm/config.cache" ]]; then
+		echo "Configuring NASM..."
+		"${WORK}/nasm-${NASM_VERSION}/configure" \
+			--quiet \
+			--config-cache \
+			--disable-werror \
+			--prefix="${SYSROOT}" \
+			--srcdir="${WORK}/nasm-${NASM_VERSION}"
+	fi
+
+	echo "Building NASM..."
+	(make -j && make install) > "${WORK}/build/nasm_build.log"
+	popd
+}
+
+function build_binutils() {
+	if [[ ! -d "${WORK}/binutils-${BINUTILS_VERSION}" ]]; then
+		echo "Downloading binutils..."
+		tarball="binutils-${BINUTILS_VERSION}.tar.gz"
+		curl -sSOL "https://ftp.gnu.org/gnu/binutils/${tarball}"
+		tar xzf "${tarball}" -C "${WORK}" && rm "${tarball}"
+	fi
+
+	mkdir -p "${WORK}/build/binutils"
+	pushd "${WORK}/build/binutils"
+
+	if [[ ! -f "${WORK}/build/binutils/config.cache" ]]; then
+		echo "Configuring binutils..."
+		"${WORK}/binutils-${BINUTILS_VERSION}/configure" \
+			--quiet \
+			--config-cache \
+			--target="${TARGET}" \
+			--prefix="${SYSROOT}" \
+			--with-sysroot="${SYSROOT}" \
+			--with-lib-path="${SYSROOT}/lib" \
+			--disable-nls \
+			--disable-werror
+	fi
+
+	echo "Building binutils..."
+	(make -j && make install) > "${WORK}/build/binutils_build.log"
+	popd
+}
+
+function build_gcc() {
+	if [[ ! -d "${WORK}/gcc-${GCC_VERSION}" ]]; then
+		echo "Downloading GCC..."
+		tarball="gcc-${GCC_VERSION}.tar.gz"
+		curl -sSOL "https://ftp.gnu.org/gnu/gcc/gcc-${GCC_VERSION}/${tarball}"
+		tar xzf "${tarball}" -C "${WORK}" && rm "${tarball}"
+
+		# no-red-zone support version of libgcc
+		echo "MULTILIB_OPTIONS += mno-red-zone" >  "${WORK}/gcc-${GCC_VERSION}/gcc/config/i386/t-${TARGET}"
+		echo "MULTILIB_DIRNAMES += no-red-zone" >> "${WORK}/gcc-${GCC_VERSION}/gcc/config/i386/t-${TARGET}"
+
+cat <<EOF >> "${WORK}/gcc-${GCC_VERSION}/gcc/config.gcc"
+case \${target} in
+	${TARGET})
+		tmake_file="\${tmake_file} i386/t-${TARGET}"
+		;;
+esac
+EOF
+	fi
+
+	mkdir -p "${WORK}/build/gcc"
+	pushd "${WORK}/build/gcc"
+
+	if [[ ! -f "${WORK}/build/gcc/config.cache" ]]; then
+		echo "Configuring GCC..."
+		"${WORK}/gcc-${GCC_VERSION}/configure" \
+			--quiet \
+			--config-cache \
+			--target="${TARGET}" \
+			--prefix="${SYSROOT}" \
+			--with-sysroot="${SYSROOT}" \
+			--with-native-system-header-dir="${SYSROOT}/include" \
+			--with-newlib \
+			--without-headers \
+			--disable-nls \
+			--enable-languages=c,c++ \
+			--disable-shared \
+			--disable-multilib \
+			--disable-decimal-float \
+			--disable-threads \
+			--disable-libatomic \
+			--disable-libgomp \
+			--disable-libmpx \
+			--disable-libquadmath \
+			--disable-libssp \
+			--disable-libvtv \
+			--disable-libstdcxx
+	fi
+
+	echo "Building GCC..."
+	(make -j all-gcc && make -j all-target-libgcc && \
+		make install-gcc && make install-target-libgcc) > "${WORK}/build/gcc_build.log"
+	popd
+}
+
+function build_libstdcxx() {
+	mkdir -p "${WORK}/build/libstdcxx"
+	pushd "${WORK}/build/libstdcxx"
+
+	if [[ ! -f "${WORK}/build/libstdcxx/config.cache" ]]; then
+		echo "Configuring libstdc++..."
+		CFLAGS="-I${SYSROOT}/include" \
+		CXXFLAGS="-I${SYSROOT}/include" \
+		"${WORK}/gcc-${GCC_VERSION}/libstdc++-v3/configure" \
+			--config-cache \
+			--host="${TARGET}" \
+			--target="${TARGET}" \
+			--prefix="${SYSROOT}" \
+			--disable-nls \
+			--disable-multilib \
+			--with-newlib \
+			--disable-libstdcxx-threads \
+			--disable-libstdcxx-pch \
+			--with-gxx-include-dir="${SYSROOT}/include/c++"
+	fi
+
+	echo "Building libstdc++..."
+	(make -j && make install) > "${WORK}/build/libstdcxx_build.log"
+	popd
+}
+
+function build_libc() {
+	if [[ ! -d "${WORK}/poplibc" ]]; then
+		echo "Downloading poplibc..."
+		git clone \
+			"https://github.com/justinian/poplibc.git" \
+			"${WORK}/poplibc"
+	else
+		echo "Updating poplibc..."
+		git -C "${WORK}/poplibc" pull
+	fi
+
+	pushd "${WORK}/poplibc"
+	echo "Building poplibc..."
+	make install PREFIX="${SYSROOT}"
+	popd
+}
+
+function update_links() {
+	for exe in `ls "${SYSROOT}/bin/${TARGET}-"*`; do
+		base=$(echo "$exe" | sed -e "s/${TARGET}-//")
+		ln -fs "${exe}" "${base}"
+	done
+}
+
+build_nasm
+build_binutils
+build_gcc
+
+update_links
+export PATH="${SYSROOT}/bin:${PATH}"
+build_libc
+build_libstdcxx

scripts/templates/build.ninja.j2

@@ -0,0 +1,86 @@
+{% extends "build.base.j2" %}
+
+{% block variables %}
+{{ super() }}
+ccflags = $ccflags $
+    -I${srcroot}/src/include $
+    -I${srcroot}/src/include/x86_64 
+{% endblock %}
+
+{% block baserules %}
+{{ super() }}
+rule makerd
+    description = Making init ramdisk
+    command = $builddir/native/makerd $in $out
+
+rule makeefi
+    description = Converting $name
+    command = objcopy $
+        -j .text $
+        -j .sdata $
+        -j .data $
+        -j .dynamic $
+        -j .dynsym $
+        -j .rel $
+        -j .rela $
+        -j .reloc $
+        --target=efi-app-x86_64 $
+        $in $out
+
+rule makefat
+    description = Creating $name
+    command = $
+        cp $srcroot/assets/diskbase.img $out; $
+        mcopy -s -D o -i $out@@1M $builddir/fatroot/* ::/
+
+rule strip
+    description = Stripping $name
+    command = $
+        cp $in $out; $
+        objcopy --only-keep-debug $out $out.debug; $
+        strip -g $out; $
+        objcopy --add-gnu-debuglink=$out.debug $out
+{% endblock %}
+
+{% block extra %}
+build $builddir/ovmf_vars.fd : cp $srcroot/assets/ovmf/x64/ovmf_vars.fd
+    name = ovmf_vars.fd
+
+build $builddir/ovmf_vars_d.fd : cp $srcroot/assets/ovmf/x64/ovmf_vars_d.fd
+    name = ovmf_vars_d.fd
+
+build $builddir/popcorn.elf | $builddir/popcorn.elf.debug : strip $builddir/host/popcorn.elf
+    name = kernel
+
+build $builddir/popcorn.dump : dump $builddir/host/popcorn.elf
+    name = kernel
+
+build $builddir/popcorn.elf-gdb.py : cp ${srcroot}/assets/debugging/popcorn.elf-gdb.py
+    name = kernel debug python scripts
+
+build $builddir/fatroot/popcorn.elf : cp $builddir/popcorn.elf
+    name = kernel to FAT image
+
+build $builddir/fatroot/efi/boot/bootx64.efi : cp $builddir/boot/boot.efi
+    name = bootloader to FAT image
+
+build $builddir/fatroot/initrd.img : makerd ${srcroot}/assets/initrd.toml | $
+    ${builddir}/native/makerd $
+    ${builddir}/user/nulldrv
+
+build $builddir/popcorn.img : makefat | $
+    $builddir/fatroot/initrd.img $
+    $builddir/fatroot/popcorn.elf $
+    $builddir/fatroot/efi/boot/bootx64.efi
+    name = popcorn.img
+
+default $
+    $builddir/ovmf_vars.fd $
+    $builddir/ovmf_vars_d.fd $
+    $builddir/popcorn.dump $
+    $builddir/popcorn.elf-gdb.py $
+    $builddir/popcorn.img
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/exe.boot.j2

@@ -0,0 +1,27 @@
+{% extends "exe.default.j2" %}
+{% block variables %}
+{{ super() }}
+
+ccflags = $ccflags $
+    -DKERNEL_FILENAME=L\"popcorn.elf\" $
+    -DGNU_EFI_USE_MS_ABI $
+    -DHAVE_USE_MS_ABI $
+    -DEFI_DEBUG=0 $
+    -DEFI_DEBUG_CLEAR_MEMORY=0 $
+    -DBOOTLOADER_DEBUG
+
+ldflags = $ldflags $
+    -T ${srcroot}/src/arch/x86_64/boot.ld $
+    -shared
+
+{% endblock %}
+
+{% block extra %}
+
+build $builddir/boot.efi : makeefi ${builddir}/{{ module.output }}
+    name = boot.efi
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/exe.kernel.j2

@@ -0,0 +1,12 @@
+{% extends "exe.default.j2" %}
+{% block variables %}
+{{ super() }}
+
+asflags = $asflags -I${srcroot}/src/kernel/
+libs = $libs
+ldflags = $ldflags -T ${srcroot}/src/arch/x86_64/kernel.ld
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/exe.tests.j2

@@ -0,0 +1,10 @@
+{% extends "exe.default.j2" %}
+{% block variables %}
+{{ super() }}
+
+ccflags = $ccflags -ggdb
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/target.boot.j2

@@ -0,0 +1,41 @@
+{% extends "target.default.j2" %}
+
+{% block binaries %}
+cc = clang
+cxx = clang++
+ld = ld
+ar = ar
+nasm = nasm
+objcopy = objcopy
+{% endblock %}
+
+{% block variables %}
+
+ccflags = $ccflags $
+	-ggdb $
+	-nostdlib $
+	-ffreestanding $
+	-nodefaultlibs $
+	-fno-builtin $
+	-mno-sse $
+	-fno-omit-frame-pointer $
+	-mno-red-zone $
+	-fshort-wchar $
+	-D__ELF__ $
+	-fPIC
+
+cxxflags = $cxxflags $
+	-fno-exceptions $
+	-fno-rtti $
+
+ldflags = $ldflags $
+	-g $
+	-nostdlib $
+	-znocombreloc $
+	-Bsymbolic $
+	-nostartfiles
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/target.host.j2

@@ -0,0 +1,45 @@
+{% extends "target.default.j2" %}
+
+{% block binaries %}
+cc = ${srcroot}/sysroot/bin/clang
+cxx = ${srcroot}/sysroot/bin/clang++
+ld = ${srcroot}/sysroot/bin/x86_64-elf-ld
+ar = ${srcroot}/sysroot/bin/x86_64-elf-ar
+nasm = ${srcroot}/sysroot/bin/nasm
+objcopy = ${srcroot}/sysroot/bin/x86_64-elf-objcopy
+{% endblock %}
+
+{% block variables %}
+
+ccflags = $ccflags $
+    -nostdlib $
+    -ffreestanding $
+    -nodefaultlibs $
+    -fno-builtin $
+    -mno-sse $
+    -fno-omit-frame-pointer $
+    -mno-red-zone $
+    -g $
+    -mcmodel=large $
+    -D__ELF__ $
+    -D__POPCORN__ $
+    -isystem${srcroot}/sysroot/include $
+    --sysroot="${srcroot}/sysroot"
+
+cxxflags = $cxxflags $
+    -fno-exceptions $
+    -fno-rtti $
+    -isystem${srcroot}/sysroot/include/c++/v1
+
+ldflags = $ldflags $
+    -g $
+    -nostdlib $
+    -znocombreloc $
+    -Bsymbolic $
+    -nostartfiles $
+    -Bstatic
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/target.native.j2

@@ -0,0 +1,15 @@
+{% extends "target.default.j2" %}
+
+{% block binaries %}
+{{ super() }}
+ld = clang++
+{% endblock %}
+
+{% block variables %}
+
+ccflags = $ccflags -g -ggdb
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/templates/target.user.j2

@@ -0,0 +1,49 @@
+{% extends "target.default.j2" %}
+
+{% block binaries %}
+cc = ${srcroot}/sysroot/bin/clang
+cxx = ${srcroot}/sysroot/bin/clang++
+ld = ${srcroot}/sysroot/bin/x86_64-elf-ld
+ar = ${srcroot}/sysroot/bin/x86_64-elf-ar
+nasm = ${srcroot}/sysroot/bin/nasm
+objcopy = ${srcroot}/sysroot/bin/x86_64-elf-objcopy
+{% endblock %}
+
+{% block variables %}
+
+ccflags = $ccflags $
+    -nostdlib $
+    -nodefaultlibs $
+    -fno-builtin $
+    -mno-sse $
+    -fno-omit-frame-pointer $
+    -mno-red-zone $
+    -g $
+    -mcmodel=large $
+    -D__ELF__ $
+    -D__POPCORN__ $
+    -isystem${srcroot}/sysroot/include $
+    --sysroot="${srcroot}/sysroot"
+
+cxxflags = $cxxflags $
+    -fno-exceptions $
+    -fno-rtti $
+    -isystem${srcroot}/sysroot/include/c++/v1
+
+ldflags = $ldflags $
+    -g $
+    -nostdlib $
+    -znocombreloc $
+    -Bsymbolic $
+    -nostartfiles $
+    -Bstatic $
+    --sysroot="${srcroot}/sysroot" $
+	-L "${srcroot}/sysroot/lib" $
+
+libs = $libs $
+	-lc
+
+{% endblock %}
+
+# vim: ft=ninja et ts=4 sts=4 sw=4
+

scripts/vmem_translate.py

@@ -0,0 +1,12 @@
+#!/usr/bin/env python
+
+def translate(i4 = 0, i3 = 0, i2 = 0, i1 = 0, offset = 0):
+    addr = (i4 << 39) + (i3 << 30) + (i2 << 21) + (i1 << 12) + offset
+    if addr & (1 << 47):
+        addr |= 0xffff000000000000
+    return addr
+
+if __name__ == "__main__":
+    import sys
+    print("{:016x}".format(translate(*map(int, sys.argv[1:]))))
+

src/arch/x86_64/kernel.ld

@@ -1,7 +1,7 @@
 ENTRY(_start)
 SECTIONS
 {
-	OFFSET = 0xFFFF800000000000;
+	OFFSET = 0xFFFFFF0000000000;
 	. = OFFSET + 0x100000;
 
 	.header : {
@@ -27,11 +27,24 @@ SECTIONS
 		*(.note.*)
 	}
 
-	.bss ALIGN(0x1000) : {
+	.bss ALIGN(16) : {
 		__bss_start = .;
 		*(.bss)
 		__bss_end = .;
 	}
 
+	.eh_frame : {
+		__eh_frame_start = .;
+		KEEP(*(.eh_frame))
+		__eh_frame_end = .;
+	}
+
+	.eh_frame_hdr : {
+		KEEP(*(.eh_frame_hdr))
+	}
+
+	__eh_frame_hdr_start = SIZEOF(.eh_frame_hdr) > 0 ? ADDR(.eh_frame_hdr) : 0;
+	__eh_frame_hdr_end = SIZEOF(.eh_frame_hdr) > 0 ? . : 0;
+
 	kernel_end = ALIGN(4096);
 }

src/boot/console.cpp

@@ -1,4 +1,3 @@
-#include <stdarg.h>
 #include <stddef.h>
 #include <stdint.h>
 
@@ -9,16 +8,55 @@
 size_t ROWS = 0;
 size_t COLS = 0;
 
-static EFI_SIMPLE_TEXT_OUT_PROTOCOL *con_out = 0;
+static EFI_SIMPLE_TEXT_OUT_PROTOCOL *m_out = 0;
 
-const CHAR16 digits[] = {u'0', u'1', u'2', u'3', u'4', u'5', u'6', u'7', u'8', u'9', u'a', u'b', u'c', u'd', u'e', u'f'};
+static const wchar_t digits[] = {u'0', u'1', u'2', u'3', u'4', u'5',
+	u'6', u'7', u'8', u'9', u'a', u'b', u'c', u'd', u'e', u'f'};
+
+console::console(EFI_SYSTEM_TABLE *system_table) :
+	m_rows(0),
+	m_cols(0),
+	m_out(nullptr)
+{
+	s_console = this;
+	m_boot = system_table->BootServices;
+	m_out = system_table->ConOut;
+}
 
 EFI_STATUS
-con_pick_mode(EFI_BOOT_SERVICES *bootsvc)
+console::initialize(const wchar_t *version)
+{
+	EFI_STATUS status;
+
+	// Might not find a video device at all, so ignore not found errors
+	status = pick_mode();
+	if (status != EFI_NOT_FOUND)
+		CHECK_EFI_STATUS_OR_FAIL(status);
+
+	status = m_out->QueryMode(m_out, m_out->Mode->Mode, &m_cols, &m_rows);
+	CHECK_EFI_STATUS_OR_RETURN(status, "QueryMode");
+
+	status = m_out->ClearScreen(m_out);
+	CHECK_EFI_STATUS_OR_RETURN(status, "ClearScreen");
+
+	m_out->SetAttribute(m_out, EFI_LIGHTCYAN);
+	m_out->OutputString(m_out, (wchar_t *)L"Popcorn loader ");
+
+	m_out->SetAttribute(m_out, EFI_LIGHTMAGENTA);
+	m_out->OutputString(m_out, (wchar_t *)version);
+
+	m_out->SetAttribute(m_out, EFI_LIGHTGRAY);
+	m_out->OutputString(m_out, (wchar_t *)L" booting...\r\n\n");
+
+	return status;
+}
+
+EFI_STATUS
+console::pick_mode()
 {
 	EFI_STATUS status;
 	EFI_GRAPHICS_OUTPUT_PROTOCOL *gfx_out_proto;
-	status = bootsvc->LocateProtocol(&guid_gfx_out, NULL, (void **)&gfx_out_proto);
+	status = m_boot->LocateProtocol(&guid_gfx_out, NULL, (void **)&gfx_out_proto);
 	CHECK_EFI_STATUS_OR_RETURN(status, "LocateProtocol gfx");
 
 	const uint32_t modes = gfx_out_proto->Mode->MaxMode;
@@ -53,106 +91,72 @@ con_pick_mode(EFI_BOOT_SERVICES *bootsvc)
 	return EFI_SUCCESS;
 }
 
-EFI_STATUS
-con_initialize(EFI_SYSTEM_TABLE *system_table, const CHAR16 *version)
-{
-	EFI_STATUS status;
-
-	EFI_BOOT_SERVICES *bootsvc = system_table->BootServices;
-	con_out = system_table->ConOut;
-
-	// Might not find a video device at all, so ignore not found errors
-	status = con_pick_mode(bootsvc);
-	if (status != EFI_NOT_FOUND)
-		CHECK_EFI_STATUS_OR_RETURN(status, "con_pick_mode");
-
-	status = con_out->QueryMode(con_out, con_out->Mode->Mode, &COLS, &ROWS);
-	CHECK_EFI_STATUS_OR_RETURN(status, "QueryMode");
-
-	status = con_out->ClearScreen(con_out);
-	CHECK_EFI_STATUS_OR_RETURN(status, "ClearScreen");
-
-	con_out->SetAttribute(con_out, EFI_LIGHTCYAN);
-	con_out->OutputString(con_out, (CHAR16 *)L"Popcorn loader ");
-
-	con_out->SetAttribute(con_out, EFI_LIGHTMAGENTA);
-	con_out->OutputString(con_out, (CHAR16 *)version);
-
-	con_out->SetAttribute(con_out, EFI_LIGHTGRAY);
-	con_out->OutputString(con_out, (CHAR16 *)L" booting...\r\n\n");
-
-	return status;
-}
-
 size_t
-con_print_hex(uint32_t n)
+console::print_hex(uint32_t n) const
 {
-	CHAR16 buffer[9];
-	CHAR16 *p = buffer;
+	wchar_t buffer[9];
+	wchar_t *p = buffer;
 	for (int i = 7; i >= 0; --i) {
 		uint8_t nibble = (n & (0xf << (i*4))) >> (i*4);
 		*p++ = digits[nibble];
 	}
 	*p = 0;
-	con_out->OutputString(con_out, buffer);
+	m_out->OutputString(m_out, buffer);
 	return 8;
 }
 
 size_t
-con_print_long_hex(uint64_t n)
+console::print_long_hex(uint64_t n) const
 {
-	CHAR16 buffer[17];
-	CHAR16 *p = buffer;
+	wchar_t buffer[17];
+	wchar_t *p = buffer;
 	for (int i = 15; i >= 0; --i) {
 		uint8_t nibble = (n & (0xf << (i*4))) >> (i*4);
 		*p++ = digits[nibble];
 	}
 	*p = 0;
-	con_out->OutputString(con_out, buffer);
+	m_out->OutputString(m_out, buffer);
 	return 16;
 }
 
 size_t
-con_print_dec(uint32_t n)
+console::print_dec(uint32_t n) const
 {
-	CHAR16 buffer[11];
-	CHAR16 *p = buffer + 10;
+	wchar_t buffer[11];
+	wchar_t *p = buffer + 10;
 	*p-- = 0;
 	do {
 		*p-- = digits[n % 10];
 		n /= 10;
 	} while (n != 0);
 
-	con_out->OutputString(con_out, ++p);
+	m_out->OutputString(m_out, ++p);
 	return 10 - (p - buffer);
 }
 
 size_t
-con_print_long_dec(uint64_t n)
+console::print_long_dec(uint64_t n) const
 {
-	CHAR16 buffer[21];
-	CHAR16 *p = buffer + 20;
+	wchar_t buffer[21];
+	wchar_t *p = buffer + 20;
 	*p-- = 0;
 	do {
 		*p-- = digits[n % 10];
 		n /= 10;
 	} while (n != 0);
 
-	con_out->OutputString(con_out, ++p);
+	m_out->OutputString(m_out, ++p);
 	return 20 - (p - buffer);
 }
 
 size_t
-con_printf(const CHAR16 *fmt, ...)
+console::vprintf(const wchar_t *fmt, va_list args) const
 {
-	CHAR16 buffer[256];
-	const CHAR16 *r = fmt;
-	CHAR16 *w = buffer;
-	va_list args;
+	wchar_t buffer[256];
+	const wchar_t *r = fmt;
+	wchar_t *w = buffer;
 	size_t count = 0;
 
-	va_start(args, fmt);
-
 	while (r && *r) {
 		if (*r != L'%') {
 			count++;
@@ -161,43 +165,43 @@ con_printf(const CHAR16 *fmt, ...)
 		}
 
 		*w = 0;
-		con_out->OutputString(con_out, buffer);
+		m_out->OutputString(m_out, buffer);
 		w = buffer;
 
 		r++; // chomp the %
 
 		switch (*r++) {
 			case L'%':
-				con_out->OutputString(con_out, L"%");
+				m_out->OutputString(m_out, const_cast<wchar_t*>(L"%"));
 				count++;
 				break;
 
 			case L'x':
-				count += con_print_hex(va_arg(args, uint32_t));
+				count += print_hex(va_arg(args, uint32_t));
 				break;
 
 			case L'd':
 			case L'u':
-				count += con_print_dec(va_arg(args, uint32_t));
+				count += print_dec(va_arg(args, uint32_t));
 				break;
 
 			case L's':
 				{
-					CHAR16 *s = va_arg(args, CHAR16*);
+					wchar_t *s = va_arg(args, wchar_t*);
 					count += wstrlen(s);
-					con_out->OutputString(con_out, s);
+					m_out->OutputString(m_out, s);
 				}
 				break;
 
 			case L'l':
 				switch (*r++) {
 					case L'x':
-						count += con_print_long_hex(va_arg(args, uint64_t));
+						count += print_long_hex(va_arg(args, uint64_t));
 						break;
 
 					case L'd':
 					case L'u':
-						count += con_print_long_dec(va_arg(args, uint64_t));
+						count += print_long_dec(va_arg(args, uint64_t));
 						break;
 
 					default:
@@ -211,44 +215,66 @@ con_printf(const CHAR16 *fmt, ...)
 	}
 
 	*w = 0;
-	con_out->OutputString(con_out, buffer);
-
-	va_end(args);
+	m_out->OutputString(m_out, buffer);
 	return count;
 }
 
-void
-con_status_begin(const CHAR16 *message)
+size_t
+console::printf(const wchar_t *fmt, ...) const
 {
-	con_out->SetAttribute(con_out, EFI_LIGHTGRAY);
-	con_out->OutputString(con_out, (CHAR16 *)message);
+	va_list args;
+	va_start(args, fmt);
+
+	size_t result = vprintf(fmt, args);
+
+	va_end(args);
+	return result;
+}
+
+size_t
+console::print(const wchar_t *fmt, ...)
+{
+	va_list args;
+	va_start(args, fmt);
+
+	size_t result = get().vprintf(fmt, args);
+
+	va_end(args);
+	return result;
 }
 
 void
-con_status_ok()
+console::status_begin(const wchar_t *message) const
 {
-	con_out->SetAttribute(con_out, EFI_LIGHTGRAY);
-	con_out->OutputString(con_out, (CHAR16 *)L"[");
-	con_out->SetAttribute(con_out, EFI_GREEN);
-	con_out->OutputString(con_out, (CHAR16 *)L"  ok  ");
-	con_out->SetAttribute(con_out, EFI_LIGHTGRAY);
-	con_out->OutputString(con_out, (CHAR16 *)L"]\r\n");
+	m_out->SetAttribute(m_out, EFI_LIGHTGRAY);
+	m_out->OutputString(m_out, (wchar_t *)message);
 }
 
 void
-con_status_fail(const CHAR16 *error)
+console::status_ok() const
 {
-	con_out->SetAttribute(con_out, EFI_LIGHTGRAY);
-	con_out->OutputString(con_out, (CHAR16 *)L"[");
-	con_out->SetAttribute(con_out, EFI_LIGHTRED);
-	con_out->OutputString(con_out, (CHAR16 *)L"failed");
-	con_out->SetAttribute(con_out, EFI_LIGHTGRAY);
-	con_out->OutputString(con_out, (CHAR16 *)L"]\r\n");
+	m_out->SetAttribute(m_out, EFI_LIGHTGRAY);
+	m_out->OutputString(m_out, (wchar_t *)L"[");
+	m_out->SetAttribute(m_out, EFI_GREEN);
+	m_out->OutputString(m_out, (wchar_t *)L"  ok  ");
+	m_out->SetAttribute(m_out, EFI_LIGHTGRAY);
+	m_out->OutputString(m_out, (wchar_t *)L"]\r\n");
+}
 
-	con_out->SetAttribute(con_out, EFI_RED);
-	con_out->OutputString(con_out, (CHAR16 *)error);
-	con_out->SetAttribute(con_out, EFI_LIGHTGRAY);
-	con_out->OutputString(con_out, (CHAR16 *)L"\r\n");
+void
+console::status_fail(const wchar_t *error) const
+{
+	m_out->SetAttribute(m_out, EFI_LIGHTGRAY);
+	m_out->OutputString(m_out, (wchar_t *)L"[");
+	m_out->SetAttribute(m_out, EFI_LIGHTRED);
+	m_out->OutputString(m_out, (wchar_t *)L"failed");
+	m_out->SetAttribute(m_out, EFI_LIGHTGRAY);
+	m_out->OutputString(m_out, (wchar_t *)L"]\r\n");
+
+	m_out->SetAttribute(m_out, EFI_RED);
+	m_out->OutputString(m_out, (wchar_t *)error);
+	m_out->SetAttribute(m_out, EFI_LIGHTGRAY);
+	m_out->OutputString(m_out, (wchar_t *)L"\r\n");
 }
 
 EFI_STATUS

src/boot/console.h

@@ -1,12 +1,38 @@
 #pragma once
+#include <stdarg.h>
 #include <stddef.h>
 #include <efi/efi.h>
 
-EFI_STATUS con_initialize(EFI_SYSTEM_TABLE *system_table, const CHAR16 *version);
-void con_status_begin(const CHAR16 *message);
-void con_status_ok();
-void con_status_fail(const CHAR16 *error);
-size_t con_printf(const CHAR16 *fmt, ...);
+class console
+{
+public:
+	console(EFI_SYSTEM_TABLE *system_table);
+
+	EFI_STATUS initialize(const wchar_t *version);
+
+	void status_begin(const wchar_t *message) const;
+	void status_fail(const wchar_t *error) const;
+	void status_ok() const;
+
+	size_t print_hex(uint32_t n) const;
+	size_t print_dec(uint32_t n) const;
+	size_t print_long_hex(uint64_t n) const;
+	size_t print_long_dec(uint64_t n) const;
+	size_t printf(const wchar_t *fmt, ...) const;
+
+	static const console & get() { return *s_console; }
+	static size_t print(const wchar_t *fmt, ...);
+
+private:
+	EFI_STATUS pick_mode();
+	size_t vprintf(const wchar_t *fmt, va_list args) const;
+
+	size_t m_rows, m_cols;
+	EFI_BOOT_SERVICES *m_boot;
+	EFI_SIMPLE_TEXT_OUT_PROTOCOL *m_out;
+
+	static console *s_console;
+};
 
 EFI_STATUS
 con_get_framebuffer(

src/boot/guids.inc

@@ -3,5 +3,9 @@ GUID(0x8868e871,0xe4f1,0x11d3,0xbc,0x22,0x00,0x80,0xc7,0x3c,0x88,0x81, guid_acpi
 GUID(0x09576e92,0x6d3f,0x11d2,0x8e,0x39,0x00,0xa0,0xc9,0x69,0x72,0x3b, guid_file_info);
 GUID(0x9042a9de,0x23dc,0x4a38,0x96,0xfb,0x7a,0xde,0xd0,0x80,0x51,0x6a, guid_gfx_out);
 GUID(0x964e5b22,0x6459,0x11d2,0x8e,0x39,0x00,0xa0,0xc9,0x69,0x72,0x3b, guid_simple_filesystem);
+GUID(0x09576e91,0x6d3f,0x11d2,0x8e,0x39,0x00,0xa0,0xc9,0x69,0x72,0x3b, guid_device_path);
+GUID(0x8b843e20,0x8132,0x4852,0x90,0xcc,0x55,0x1a,0x4e,0x4a,0x7f,0x1c, guid_device_path_to_text);
+
+GUID(0x10d0669c,0x9ec6,0x4268,0xbc,0x48,0xff,0x74,0x75,0x21,0xfe,0x07, guid_popcorn_vendor);
 
 // vim: ft=c

src/boot/loader.cpp

@@ -6,8 +6,59 @@
 
 #define PAGE_SIZE 0x1000
 
-static CHAR16 kernel_name[] = KERNEL_FILENAME;
-static CHAR16 font_name[] = KERNEL_FONT;
+static wchar_t kernel_name[] = KERNEL_FILENAME;
+static wchar_t initrd_name[] = INITRD_FILENAME;
+
+EFI_STATUS
+loader_alloc_aligned(
+	EFI_BOOT_SERVICES *bootsvc,
+	EFI_MEMORY_TYPE mem_type,
+	size_t *length,
+	void **pages)
+{
+	EFI_STATUS status;
+	EFI_PHYSICAL_ADDRESS addr;
+
+	size_t alignment = PAGE_SIZE;
+	while (alignment < *length)
+		alignment *= 2;
+
+	size_t page_count = alignment / PAGE_SIZE;
+	*length = alignment;
+
+	con_debug(L"Trying to find %d aligned pages for %x\n", page_count, mem_type);
+
+	status = bootsvc->AllocatePages(AllocateAnyPages, mem_type, page_count * 2, &addr);
+	CHECK_EFI_STATUS_OR_RETURN(status, L"Allocating %d pages for alignment", page_count * 2);
+	con_debug(L"    Found %d pages at %lx\n", page_count * 2, addr);
+
+	EFI_PHYSICAL_ADDRESS aligned = addr;
+	aligned = ((aligned - 1) & ~(alignment - 1)) + alignment;
+	con_debug(L"    Aligning %lx to %lx\n", addr, aligned);
+
+	size_t before = 
+		(reinterpret_cast<uint64_t>(aligned) -
+		reinterpret_cast<uint64_t>(addr)) /
+		PAGE_SIZE;
+
+	if (before) {
+		con_debug(L"    Freeing %d initial pages\n", before);
+		bootsvc->FreePages(addr, before);
+	}
+
+	size_t after = page_count - before;
+	if (after) {
+		EFI_PHYSICAL_ADDRESS end = 
+			reinterpret_cast<EFI_PHYSICAL_ADDRESS>(
+				reinterpret_cast<uint64_t>(aligned) +
+				page_count * PAGE_SIZE);
+		con_debug(L"    Freeing %d remaining pages\n", after);
+		bootsvc->FreePages(end, after);
+	}
+
+	*pages = (void *)aligned;
+	return EFI_SUCCESS;
+}
 
 EFI_STATUS
 loader_alloc_pages(
@@ -21,12 +72,10 @@ loader_alloc_pages(
 	size_t page_count = ((*length - 1) / PAGE_SIZE) + 1;
 	EFI_PHYSICAL_ADDRESS addr = (EFI_PHYSICAL_ADDRESS)*pages;
 
+	con_debug(L"Trying to find %d non-aligned pages for %x at %lx\n",
+			page_count, mem_type, addr);
+
 	status = bootsvc->AllocatePages(AllocateAddress, mem_type, page_count, &addr);
-	if (status == EFI_NOT_FOUND || status == EFI_OUT_OF_RESOURCES) {
-		// couldn't get the address we wanted, try loading the kernel anywhere
-		status =
-			bootsvc->AllocatePages(AllocateAnyPages, mem_type, page_count, &addr);
-	}
 	CHECK_EFI_STATUS_OR_RETURN(status,
 		L"Allocating %d kernel pages type %x",
 		page_count, mem_type);
@@ -38,7 +87,7 @@ loader_alloc_pages(
 }
 
 EFI_STATUS
-loader_load_font(
+loader_load_initrd(
 	EFI_BOOT_SERVICES *bootsvc,
 	EFI_FILE_PROTOCOL *root,
 	struct loader_data *data)
@@ -46,13 +95,13 @@ loader_load_font(
 	EFI_STATUS status;
 
 	EFI_FILE_PROTOCOL *file = NULL;
-	status = root->Open(root, &file, (CHAR16 *)font_name, EFI_FILE_MODE_READ,
+	status = root->Open(root, &file, (wchar_t *)initrd_name, EFI_FILE_MODE_READ,
 						EFI_FILE_READ_ONLY | EFI_FILE_HIDDEN | EFI_FILE_SYSTEM);
 
 	if (status == EFI_NOT_FOUND)
 		return status;
 
-	CHECK_EFI_STATUS_OR_RETURN(status, L"Opening file %s", font_name);
+	CHECK_EFI_STATUS_OR_RETURN(status, L"Opening file %s", initrd_name);
 
 	char info[sizeof(EFI_FILE_INFO) + 100];
 	size_t info_length = sizeof(info);
@@ -60,16 +109,16 @@ loader_load_font(
 	status = file->GetInfo(file, &guid_file_info, &info_length, info);
 	CHECK_EFI_STATUS_OR_RETURN(status, L"Getting file info");
 
-	data->font_length = ((EFI_FILE_INFO *)info)->FileSize;
+	data->initrd_length = ((EFI_FILE_INFO *)info)->FileSize;
 
-	status = loader_alloc_pages(
+	status = loader_alloc_aligned(
 			bootsvc,
-			KERNEL_FONT_MEMTYPE,
-			&data->font_length,
-			&data->font);
+			memtype_initrd,
+			&data->initrd_length,
+			&data->initrd);
 	CHECK_EFI_STATUS_OR_RETURN(status, L"Allocating pages");
 
-	status = file->Read(file, &data->font_length, data->font);
+	status = file->Read(file, &data->initrd_length, data->initrd);
 	CHECK_EFI_STATUS_OR_RETURN(status, L"Reading file");
 
 	status = file->Close(file);
@@ -87,8 +136,10 @@ loader_load_elf(
 {
 	EFI_STATUS status;
 
+	con_debug(L"Opening kernel file %s\r\n", (wchar_t *)kernel_name);
+
 	EFI_FILE_PROTOCOL *file = NULL;
-	status = root->Open(root, &file, (CHAR16 *)kernel_name, EFI_FILE_MODE_READ,
+	status = root->Open(root, &file, (wchar_t *)kernel_name, EFI_FILE_MODE_READ,
 						EFI_FILE_READ_ONLY | EFI_FILE_HIDDEN | EFI_FILE_SYSTEM);
 
 	if (status == EFI_NOT_FOUND)
@@ -107,6 +158,8 @@ loader_load_elf(
 	status = file->Read(file, &length, &header);
 	CHECK_EFI_STATUS_OR_RETURN(status, L"Reading ELF header");
 
+	con_debug(L"Read %u bytes of ELF header\r\n", length);
+
 	if (length < sizeof(struct elf_header))
 		CHECK_EFI_STATUS_OR_RETURN(EFI_LOAD_ERROR, L"Incomplete read of ELF header");
 
@@ -131,10 +184,13 @@ loader_load_elf(
 		header.machine != 0x3e)
 		CHECK_EFI_STATUS_OR_RETURN(EFI_LOAD_ERROR, L"ELF load error: wrong machine architecture");
 
+	con_debug(L"ELF is valid, entrypoint %lx\r\n", header.entrypoint);
+
 	data->kernel_entry = (void *)header.entrypoint;
 
 	struct elf_program_header prog_header;
 	for (int i = 0; i < header.ph_num; ++i) {
+
 		status = file->SetPosition(file, header.ph_offset + i * header.ph_entsize);
 		CHECK_EFI_STATUS_OR_RETURN(status, L"Setting ELF file position");
 
@@ -146,13 +202,14 @@ loader_load_elf(
 
 		length = prog_header.mem_size;
 		void *addr = (void *)(prog_header.vaddr - KERNEL_VIRT_ADDRESS);
-		status = loader_alloc_pages(bootsvc, KERNEL_MEMTYPE, &length, &addr);
+		status = loader_alloc_pages(bootsvc, memtype_kernel, &length, &addr);
 		CHECK_EFI_STATUS_OR_RETURN(status, L"Allocating kernel pages");
 
 		if (data->kernel == 0)
 			data->kernel = addr;
 		data->kernel_length = (uint64_t)addr + length - (uint64_t)data->kernel;
 	}
+	con_debug(L"Read %d ELF program headers\r\n", header.ph_num);
 
 	struct elf_section_header sec_header;
 	for (int i = 0; i < header.sh_num; ++i) {
@@ -163,7 +220,9 @@ loader_load_elf(
 		status = file->Read(file, &length, &sec_header);
 		CHECK_EFI_STATUS_OR_RETURN(status, L"Reading ELF section header");
 
-		if ((sec_header.flags & ELF_SHF_ALLOC) == 0) continue;
+		if ((sec_header.flags & ELF_SHF_ALLOC) == 0) {
+			continue;
+		}
 
 		void *addr = (void *)(sec_header.addr - KERNEL_VIRT_ADDRESS);
 
@@ -178,6 +237,7 @@ loader_load_elf(
 			bootsvc->SetMem(addr, sec_header.size, 0);
 		}
 	}
+	con_debug(L"Read %d ELF section headers\r\n", header.ph_num);
 
 	status = file->Close(file);
 	CHECK_EFI_STATUS_OR_RETURN(status, L"Closing file handle");
@@ -215,21 +275,21 @@ loader_load_kernel(
 		if (status == EFI_NOT_FOUND)
 			continue;
 
-		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_file: %s", kernel_name);
+		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_elf: %s", kernel_name);
 
-		data->font = (void *)((uint64_t)data->kernel + data->kernel_length);
-		status = loader_load_font(bootsvc, root, data);
-
-		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_file: %s", font_name);
-
-		data->data = (void *)((uint64_t)data->font + data->font_length);
+		data->data = (void *)((uint64_t)data->kernel + data->kernel_length);
 		data->data_length += PAGE_SIZE; // extra page for map growth
-		status = loader_alloc_pages(
+
+		status = loader_alloc_aligned(
 				bootsvc,
-				KERNEL_DATA_MEMTYPE,
+				memtype_data,
 				&data->data_length,
 				&data->data);
-		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_alloc_pages: kernel data");
+		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_alloc_aligned: kernel data");
+
+		data->initrd = (void *)((uint64_t)data->data + data->data_length);
+		status = loader_load_initrd(bootsvc, root, data);
+		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_file: %s", initrd_name);
 
 		return EFI_SUCCESS;
 	}

src/boot/loader.h

@@ -9,35 +9,15 @@
 #endif
 
 #ifndef KERNEL_VIRT_ADDRESS
-#define KERNEL_VIRT_ADDRESS 0xFFFF800000000000
-#endif
-
-#ifndef VIRTUAL_OFFSET
-#define VIRTUAL_OFFSET 0xf00000000
-#endif
-
-#ifndef KERNEL_MEMTYPE
-#define KERNEL_MEMTYPE 0x80000000
-#endif
-
-#ifndef KERNEL_FONT_MEMTYPE
-#define KERNEL_FONT_MEMTYPE 0x80000001
-#endif
-
-#ifndef KERNEL_DATA_MEMTYPE
-#define KERNEL_DATA_MEMTYPE 0x80000002
-#endif
-
-#ifndef KERNEL_PT_MEMTYPE
-#define KERNEL_PT_MEMTYPE 0x80000004
+#define KERNEL_VIRT_ADDRESS 0xFFFFFF0000000000
 #endif
 
 #ifndef KERNEL_FILENAME
 #define KERNEL_FILENAME L"kernel.elf"
 #endif
 
-#ifndef KERNEL_FONT
-#define KERNEL_FONT L"screenfont.psf"
+#ifndef INITRD_FILENAME
+#define INITRD_FILENAME L"initrd.img"
 #endif
 
 struct loader_data {
@@ -45,8 +25,8 @@ struct loader_data {
 	void *kernel_entry;
 	size_t kernel_length;
 
-	void *font;
-	size_t font_length;
+	void *initrd;
+	size_t initrd_length;
 
 	void *data;
 	size_t data_length;

src/boot/main.cpp

@@ -5,11 +5,15 @@
 
 #include "console.h"
 #include "guids.h"
-#include "kernel_data.h"
+#include "kernel_args.h"
 #include "loader.h"
 #include "memory.h"
 #include "utility.h"
 
+#ifndef SCRATCH_PAGES
+#define SCRATCH_PAGES 64
+#endif
+
 #ifndef GIT_VERSION_WIDE
 #define GIT_VERSION_WIDE L"no version"
 #endif
@@ -30,20 +34,68 @@ struct kernel_header {
 };
 #pragma pack(pop)
 
+using kernel_entry = void (*)(kernel_args *);
+
+static void
+type_to_wchar(wchar_t *into, uint32_t type)
+{
+	for (int j=0; j<4; ++j)
+		into[j] = static_cast<wchar_t>(reinterpret_cast<char *>(&type)[j]);
+}
+
 EFI_STATUS
+detect_debug_mode(EFI_RUNTIME_SERVICES *run, kernel_args *header) {
+	wchar_t var_name[] = L"debug";
+
+	EFI_STATUS status;
+	uint8_t debug = 0;
+	UINTN var_size = sizeof(debug);
+
+#ifdef __POPCORN_SET_DEBUG_UEFI_VAR__
+	debug = __POPCORN_SET_DEBUG_UEFI_VAR__;
+	uint32_t attrs =
+		EFI_VARIABLE_NON_VOLATILE |
+		EFI_VARIABLE_BOOTSERVICE_ACCESS |
+		EFI_VARIABLE_RUNTIME_ACCESS;
+	status = run->SetVariable(
+			var_name,
+			&guid_popcorn_vendor,
+			attrs,
+			var_size,
+			&debug);
+	CHECK_EFI_STATUS_OR_RETURN(status, "detect_debug_mode::SetVariable");
+#endif
+
+	status = run->GetVariable(
+			var_name,
+			&guid_popcorn_vendor,
+			nullptr,
+			&var_size,
+			&debug);
+	CHECK_EFI_STATUS_OR_RETURN(status, "detect_debug_mode::GetVariable");
+
+	if (debug)
+		header->flags |= POPCORN_FLAG_DEBUG;
+
+	return EFI_SUCCESS;
+}
+
+extern "C" EFI_STATUS
 efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 {
 	EFI_STATUS status;
 	EFI_BOOT_SERVICES *bootsvc = system_table->BootServices;
 	EFI_RUNTIME_SERVICES *runsvc = system_table->RuntimeServices;
 
+	console con(system_table);
+
 	// When checking console initialization, use CHECK_EFI_STATUS_OR_RETURN
 	// because we can't be sure if the console was fully set up
-	status = con_initialize(system_table, GIT_VERSION_WIDE);
-	CHECK_EFI_STATUS_OR_RETURN(status, "con_initialize");
+	status = con.initialize(GIT_VERSION_WIDE);
+	CHECK_EFI_STATUS_OR_RETURN(status, "console::initialize");
 	// From here on out, we can use CHECK_EFI_STATUS_OR_FAIL instead
 
-	memory_init_pointer_fixup(bootsvc, runsvc);
+	memory_init_pointer_fixup(bootsvc, runsvc, SCRATCH_PAGES);
 
 	// Find ACPI tables. Ignore ACPI 1.0 if a 2.0 table is found.
 	//
@@ -65,55 +117,56 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 	status = memory_get_map_length(bootsvc, &data_length);
 	CHECK_EFI_STATUS_OR_FAIL(status);
 
-	size_t header_size = sizeof(struct popcorn_data);
-	const size_t header_align = alignof(struct popcorn_data);
+	size_t header_size = sizeof(kernel_args);
+	const size_t header_align = alignof(kernel_args);
 	if (header_size % header_align)
 		header_size += header_align - (header_size % header_align);
 
 	data_length += header_size;
 
-
 	// Load the kernel image from disk and check it
 	//
-	con_printf(L"Loading kernel into memory...\r\n");
+	console::print(L"Loading kernel into memory...\r\n");
 
 	struct loader_data load;
 	load.data_length = data_length;
 	status = loader_load_kernel(bootsvc, &load);
 	CHECK_EFI_STATUS_OR_FAIL(status);
 
-	con_printf(L"    %u image bytes at 0x%x\r\n", load.kernel_length, load.kernel);
-	con_printf(L"    %u font bytes at 0x%x\r\n", load.font_length, load.font);
-	con_printf(L"    %u data bytes at 0x%x\r\n", load.data_length, load.data);
+	console::print(L"    %x image bytes  at 0x%x\r\n", load.kernel_length, load.kernel);
+	console::print(L"    %x data bytes   at 0x%x\r\n", load.data_length, load.data);
+	console::print(L"    %x initrd bytes at 0x%x\r\n", load.initrd_length, load.initrd);
 
 	struct kernel_header *version = (struct kernel_header *)load.kernel;
 	if (version->magic != KERNEL_HEADER_MAGIC) {
-		con_printf(L"    bad magic %x\r\n", version->magic);
+		console::print(L"    bad magic %x\r\n", version->magic);
 		CHECK_EFI_STATUS_OR_FAIL(EFI_CRC_ERROR);
 	}
 
-	con_printf(L"    Kernel version %d.%d.%d %x%s\r\n",
+	console::print(L"    Kernel version %d.%d.%d %x%s\r\n",
 			version->major, version->minor, version->patch, version->gitsha & 0x0fffffff,
-			version->gitsha & 0xf0000000 ? "*" : "");
-	con_printf(L"    Entrypoint 0x%x\r\n", load.kernel_entry);
+			version->gitsha & 0xf0000000 ? L"*" : L"");
+	console::print(L"    Entrypoint 0x%x\r\n", load.kernel_entry);
 
-	void (*kernel_main)() = load.kernel_entry;
+	kernel_entry kernel_main =
+		reinterpret_cast<kernel_entry>(load.kernel_entry);
 	memory_mark_pointer_fixup((void **)&kernel_main);
 
 	// Set up the kernel data pages to pass to the kernel
 	//
-	struct popcorn_data *data_header = (struct popcorn_data *)load.data; 
+	struct kernel_args *data_header = (struct kernel_args *)load.data;
 	memory_mark_pointer_fixup((void **)&data_header);
 
 	data_header->magic = DATA_HEADER_MAGIC;
 	data_header->version = DATA_HEADER_VERSION;
-	data_header->length = sizeof(struct popcorn_data);
+	data_header->length = sizeof(struct kernel_args);
 
+	data_header->scratch_pages = SCRATCH_PAGES;
 	data_header->flags = 0;
 
-	data_header->font = load.font;
-	data_header->font_length = load.font_length;
-	memory_mark_pointer_fixup((void **)&data_header->font);
+	data_header->initrd = load.initrd;
+	data_header->initrd_length = load.initrd_length;
+	memory_mark_pointer_fixup((void **)&data_header->initrd);
 
 	data_header->data = load.data;
 	data_header->data_length = load.data_length;
@@ -148,8 +201,9 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 	// Save the memory map and tell the firmware we're taking control.
 	//
 	struct memory_map map;
-	map.entries = data_header->memory_map;
 	map.length = (load.data_length - header_size);
+	map.entries =
+		reinterpret_cast<EFI_MEMORY_DESCRIPTOR *>(data_header->memory_map);
 
 	status = memory_get_map(bootsvc, &map);
 	CHECK_EFI_STATUS_OR_FAIL(status);
@@ -157,6 +211,8 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 	data_header->memory_map_length = map.length;
 	data_header->memory_map_desc_size = map.size;
 
+	detect_debug_mode(runsvc, data_header);
+
 	// bootsvc->Stall(5000000);
 
 	status = bootsvc->ExitBootServices(image_handle, map.key);

src/boot/memory.cpp

@@ -4,13 +4,18 @@
 #include "memory.h"
 #include "utility.h"
 
+const EFI_MEMORY_TYPE memtype_kernel  = static_cast<EFI_MEMORY_TYPE>(0x80000000);
+const EFI_MEMORY_TYPE memtype_data    = static_cast<EFI_MEMORY_TYPE>(0x80000001);
+const EFI_MEMORY_TYPE memtype_initrd  = static_cast<EFI_MEMORY_TYPE>(0x80000002);
+const EFI_MEMORY_TYPE memtype_scratch = static_cast<EFI_MEMORY_TYPE>(0x80000003);
+
 #define INCREMENT_DESC(p, b)  (EFI_MEMORY_DESCRIPTOR*)(((uint8_t*)(p))+(b))
 
 size_t fixup_pointer_index = 0;
 void **fixup_pointers[64];
 uint64_t *new_pml4 = 0;
 
-const CHAR16 *memory_type_names[] = {
+const wchar_t *memory_type_names[] = {
 	L"EfiReservedMemoryType",
 	L"EfiLoaderCode",
 	L"EfiLoaderData",
@@ -28,13 +33,17 @@ const CHAR16 *memory_type_names[] = {
 	L"EfiPersistentMemory",
 };
 
-static const CHAR16 *
+static const wchar_t *
 memory_type_name(UINT32 value)
 {
-	if (value >= (sizeof(memory_type_names) / sizeof(CHAR16 *))) {
-		if (value == KERNEL_DATA_MEMTYPE) return L"Kernel Data";
-		else if (value == KERNEL_MEMTYPE) return L"Kernel Image";
-		else return L"Bad Type Value";
+	if (value >= (sizeof(memory_type_names) / sizeof(wchar_t *))) {
+		switch (value) {
+			case memtype_kernel:  return L"Kernel Data";
+			case memtype_data:    return L"Kernel Data";
+			case memtype_initrd:  return L"Initial Ramdisk";
+			case memtype_scratch: return L"Kernel Scratch Space";
+			default: return L"Bad Type Value";
+		}
 	}
 	return memory_type_names[value];
 }
@@ -50,7 +59,7 @@ memory_update_marked_addresses(EFI_EVENT UNUSED *event, void *context)
 }
 
 EFI_STATUS
-memory_init_pointer_fixup(EFI_BOOT_SERVICES *bootsvc, EFI_RUNTIME_SERVICES *runsvc)
+memory_init_pointer_fixup(EFI_BOOT_SERVICES *bootsvc, EFI_RUNTIME_SERVICES *runsvc, unsigned scratch_pages)
 {
 	EFI_STATUS status;
 	EFI_EVENT event;
@@ -67,7 +76,7 @@ memory_init_pointer_fixup(EFI_BOOT_SERVICES *bootsvc, EFI_RUNTIME_SERVICES *runs
 	// Reserve a page for our replacement PML4, plus some pages for the kernel to use
 	// as page tables while it gets started.
 	EFI_PHYSICAL_ADDRESS addr = 0;
-	status = bootsvc->AllocatePages(AllocateAnyPages, EfiLoaderData, 16, &addr);
+	status = bootsvc->AllocatePages(AllocateAnyPages, memtype_scratch, scratch_pages, &addr);
 	CHECK_EFI_STATUS_OR_RETURN(status, "Failed to allocate page table pages.");
 
 	new_pml4 = (uint64_t *)addr;
@@ -139,19 +148,19 @@ memory_dump_map(struct memory_map *map)
 
 	const size_t count = map->length / map->size;
 
-	con_printf(L"Memory map:\n");
-	con_printf(L"\t	Descriptor Count: %d (%d bytes)\n", count, map->length);
-	con_printf(L"\t  Descriptor Size: %d bytes\n", map->size);
-	con_printf(L"\t      Type offset: %d\n\n", offsetof(EFI_MEMORY_DESCRIPTOR, Type));
+	console::print(L"Memory map:\n");
+	console::print(L"\t	Descriptor Count: %d (%d bytes)\n", count, map->length);
+	console::print(L"\t  Descriptor Size: %d bytes\n", map->size);
+	console::print(L"\t      Type offset: %d\n\n", offsetof(EFI_MEMORY_DESCRIPTOR, Type));
 
 	EFI_MEMORY_DESCRIPTOR *end = INCREMENT_DESC(map->entries, map->length);
 	EFI_MEMORY_DESCRIPTOR *d = map->entries;
 	while (d < end) {
 		int runtime = (d->Attribute & EFI_MEMORY_RUNTIME) == EFI_MEMORY_RUNTIME;
-		con_printf(L"%s%s ", memory_type_name(d->Type), runtime ? L"*" : L" ");
-		con_printf(L"%lx ", d->PhysicalStart);
-		con_printf(L"%lx ", d->VirtualStart);
-		con_printf(L"[%4d]\n", d->NumberOfPages);
+		console::print(L"%s%s ", memory_type_name(d->Type), runtime ? L"*" : L" ");
+		console::print(L"%lx ", d->PhysicalStart);
+		console::print(L"%lx ", d->VirtualStart);
+		console::print(L"[%4d]\n", d->NumberOfPages);
 
 		d = INCREMENT_DESC(d, map->size);
 	}
@@ -187,10 +196,12 @@ memory_virtualize(EFI_RUNTIME_SERVICES *runsvc, struct memory_map *map)
 	EFI_MEMORY_DESCRIPTOR *d = map->entries;
 	while (d < end) {
 		switch (d->Type) {
-		case KERNEL_MEMTYPE:
-		case KERNEL_FONT_MEMTYPE:
-		case KERNEL_DATA_MEMTYPE:
+		case memtype_kernel:
+		case memtype_data:
+		case memtype_initrd:
+		case memtype_scratch:
 			d->Attribute |= EFI_MEMORY_RUNTIME;
+			d->VirtualStart = d->PhysicalStart + KERNEL_VIRT_ADDRESS;
 
 		default:
 			if (d->Attribute & EFI_MEMORY_RUNTIME) {

src/boot/memory.h

@@ -1,6 +1,11 @@
 #pragma once
 #include <efi/efi.h>
 
+extern const EFI_MEMORY_TYPE memtype_kernel;
+extern const EFI_MEMORY_TYPE memtype_data;
+extern const EFI_MEMORY_TYPE memtype_initrd;
+extern const EFI_MEMORY_TYPE memtype_scratch;
+
 struct memory_map {
 	size_t length;
 	size_t size;
@@ -9,7 +14,10 @@ struct memory_map {
 	EFI_MEMORY_DESCRIPTOR *entries;
 };
 
-EFI_STATUS memory_init_pointer_fixup(EFI_BOOT_SERVICES *bootsvc, EFI_RUNTIME_SERVICES *runsvc);
+EFI_STATUS memory_init_pointer_fixup(
+		EFI_BOOT_SERVICES *bootsvc,
+		EFI_RUNTIME_SERVICES *runsvc,
+		unsigned scratch_pages);
 void memory_mark_pointer_fixup(void **p);
 
 EFI_STATUS memory_get_map_length(EFI_BOOT_SERVICES *bootsvc, size_t *size);

src/boot/reloc.cpp

@@ -38,6 +38,7 @@
 #include <efi/efi.h>
 #include <elf.h>
 
+extern "C"
 EFI_STATUS _relocate (long ldbase, Elf64_Dyn *dyn,
 		      EFI_HANDLE image EFI_UNUSED,
 		      EFI_SYSTEM_TABLE *systab EFI_UNUSED)

src/boot/utility.cpp

@@ -2,7 +2,7 @@
 
 struct error_code_desc {
 	EFI_STATUS code;
-	CHAR16 *name;
+	const wchar_t *name;
 };
 
 // Based off the gnu-efi table
@@ -47,7 +47,7 @@ struct error_code_desc error_table[] = {
 	{ 0, NULL }
 };
 
-const CHAR16 *
+const wchar_t *
 util_error_message(EFI_STATUS status)
 {
 	int32_t i = -1;
@@ -62,7 +62,7 @@ util_error_message(EFI_STATUS status)
 }
 
 size_t
-wstrlen(const CHAR16 *s)
+wstrlen(const wchar_t *s)
 {
 	size_t count = 0;
 	while (s && *s++) count++;

src/boot/utility.h

@@ -5,18 +5,18 @@
 
 #define UNUSED __attribute__((unused))
 
-size_t wstrlen(const CHAR16 *s);
-const CHAR16 *util_error_message(EFI_STATUS status);
+size_t wstrlen(const wchar_t *s);
+const wchar_t *util_error_message(EFI_STATUS status);
 
 #define CHECK_EFI_STATUS_OR_RETURN(s, msg, ...) \
 	if (EFI_ERROR((s))) { \
-		con_printf(L"ERROR: " msg L": %s\r\n", ##__VA_ARGS__, util_error_message(s)); \
+		console::print(L"ERROR: " msg L": %s\r\n", ##__VA_ARGS__, util_error_message(s)); \
 		return (s); \
 	}
 
 #define CHECK_EFI_STATUS_OR_FAIL(s) \
 	if (EFI_ERROR((s))) { \
-		con_status_fail(util_error_message(s)); \
+		console::get().status_fail(util_error_message(s)); \
 		while (1) __asm__("hlt"); \
 	}
 
@@ -32,3 +32,9 @@ const CHAR16 *util_error_message(EFI_STATUS status);
 			: "r"((uint64_t)s), "r"((uint64_t)d), "r"((uint64_t)__LINE__) \
 			: "rax", "rdx", "r8", "r9", "r10"); \
 	}
+
+#ifdef BOOTLOADER_DEBUG
+#define con_debug(...) console::print(L"DEBUG: " __VA_ARGS__)
+#else
+#define con_debug(...)
+#endif

src/boot/wscript

@@ -1,61 +0,0 @@
-
-def configure(ctx):
-    from os.path import join
-
-    lds_path = join(ctx.env.ARCH_D, "boot.ld")
-
-    ctx.env.append_value('DEFINES_EFI', [
-        'KERNEL_FILENAME=L"{}"'.format(ctx.env.KERNEL_FILENAME),
-        'GNU_EFI_USE_MS_ABI',
-        'HAVE_USE_MS_ABI',
-        'EFI_DEBUG=0',
-        'EFI_DEBUG_CLEAR_MEMORY=0',
-    ])
-    ctx.env.append_value('CFLAGS_EFI', ['-fPIC', '-fshort-wchar'])
-    ctx.env.append_value('LINKFLAGS_EFI', [
-        '-shared',
-        '-T', lds_path,
-    ])
-    ctx.env.append_value('SECTIONS_EFI', [
-    ])
-
-def build(bld):
-    sources = bld.path.ant_glob("**/*.c")
-    sources += bld.path.ant_glob("**/*.s")
-
-    bld.program(
-        source      = sources,
-        target      = "boot.elf",
-        use         = 'EFI',
-    )
-
-    from waflib.Task import Task
-    class make_efi(Task):
-        color = 'YELLOW'
-        def keyword(self):
-            return "Creating"
-        def __str__(self):
-            node = self.outputs[0]
-            return node.path_from(node.ctx.launch_node())
-        def run(self):
-            from subprocess import check_call as call
-
-            args = self.env.objcopy
-            sections = [".text", ".sdata", ".data", ".dynamic",
-                        ".dynsym", ".rel", ".rela", ".reloc"]
-            for s in sections: args.extend(['-j', s])
-            args.append('--target=efi-app-' + self.env.POPCORN_ARCH)
-            args.append(self.inputs[0].abspath())
-            args.append(self.outputs[0].abspath())
-            call(args)
-
-    src = bld.path
-    out = bld.path.get_bld()
-
-    efi = make_efi(env=bld.env)
-    efi.set_inputs([out.make_node("boot.elf")])
-    efi.set_outputs([out.make_node("boot.efi")])
-    bld.add_to_group(efi)
-
-
-# vim: ft=python et sw=4

src/drivers/ahci/driver.cpp

@@ -3,13 +3,15 @@
 #include "log.h"
 #include "pci.h"
 
+namespace ahci {
 
-ahci_driver::ahci_driver()
+
+driver::driver()
 {
 }
 
 void
-ahci_driver::register_device(pci_device *device)
+driver::register_device(pci_device *device)
 {
 	log::info(logs::driver, "AHCI registering device %d:%d:%d:",
 			device->bus(), device->device(), device->function());
@@ -17,12 +19,4 @@ ahci_driver::register_device(pci_device *device)
 	ahci::hba &hba = m_devices.emplace(device);
 }
 
-ahci::port *
-ahci_driver::find_disk()
-{
-	for (auto &hba : m_devices) {
-		ahci::port *d = hba.find_disk();
-		if (d) return d;
-	}
-	return nullptr;
-}
+} // namespace

src/drivers/ahci/driver.h

@@ -6,13 +6,15 @@
 
 class pci_device;
 
+namespace ahci {
+
 
 /// Basic AHCI driver
-class ahci_driver
+class driver
 {
 public:
 	/// Constructor.
-	ahci_driver();
+	driver();
 
 	/// Register a device with the driver
 	/// \arg device  The PCI device to handle
@@ -22,10 +24,8 @@ public:
 	/// \arg device  The PCI device to remove
 	void unregister_device(pci_device *device);
 
-	/// Debug: find the first disk
-	ahci::port * find_disk();
-
 private:
 	kutil::vector<ahci::hba> m_devices;
 };
 
+} // namespace

src/drivers/ahci/hba.cpp

@@ -1,7 +1,9 @@
 #include <stdint.h>
 #include "ahci/ata.h"
 #include "ahci/hba.h"
+#include "console.h"
 #include "device_manager.h"
+#include "fs/gpt.h"
 #include "log.h"
 #include "page_manager.h"
 #include "pci.h"
@@ -57,10 +59,14 @@ void irq_cb(void *data)
 hba::hba(pci_device *device)
 {
 	page_manager *pm = page_manager::get();
+	device_manager &dm = device_manager::get();
 
 	uint32_t bar5 = device->get_bar(5);
-	m_data = reinterpret_cast<hba_data *>(bar5 & ~0xfffull);
-	pm->map_offset_pointer(reinterpret_cast<void **>(&m_data), 0x2000);
+	log::debug(logs::driver, "HBA raw BAR5 is %08lx", bar5);
+
+	void *data = reinterpret_cast<void *>(bar5 & ~0xfffull);
+	pm->map_offset_pointer(&data, 0x2000);
+	m_data = reinterpret_cast<hba_data volatile *>(data);
 
 	if (! bitfield_has(m_data->cap, hba_cap::ahci_only))
 		m_data->host_control |= 0x80000000; // Enable AHCI mode
@@ -70,17 +76,17 @@ hba::hba(pci_device *device)
 	unsigned ports = (icap & 0xf) + 1;
 	unsigned slots = ((icap >> 8) & 0x1f) + 1;
 
-	log::debug(logs::driver, "  %d ports", ports);
+	log::debug(logs::driver, "  %d ports: %08x", ports, m_data->port_impl);
 	log::debug(logs::driver, "  %d command slots", slots);
 
-	port_data *pd = reinterpret_cast<port_data *>(
+	auto *pd = reinterpret_cast<port_data volatile *>(
 			kutil::offset_pointer(m_data, 0x100));
 
 	bool needs_interrupt = false;
 	m_ports.ensure_capacity(ports);
 	for (unsigned i = 0; i < ports; ++i) {
 		bool impl = ((m_data->port_impl & (1 << i)) != 0);
-		port &p = m_ports.emplace(i, kutil::offset_pointer(pd, 0x80 * i), impl);
+		port &p = m_ports.emplace(this, i, kutil::offset_pointer(pd, 0x80 * i), impl);
 		if (p.get_state() == port::state::active)
 			needs_interrupt = true;
 	}
@@ -89,29 +95,48 @@ hba::hba(pci_device *device)
 		device_manager::get().allocate_msi("AHCI Device", *device, irq_cb, this);
 		m_data->host_control |= 0x02; // enable interrupts
 	}
-}
 
-port *
-hba::find_disk()
-{
-	for (auto &port : m_ports) {
-		if (port.get_state() == port::state::active &&
-			port.get_type() == sata_signature::sata_drive)
-			return &port;
+	for (auto &p : m_ports) {
+		if (!p.active()) continue;
+
+		if (p.get_type() == sata_signature::sata_drive) {
+			p.sata_reconnect();
+			/*
+			if (fs::partition::load(&p) == 0)
+				dm.register_block_device(&p);
+			*/
+		}
 	}
-
-	return nullptr;
 }
 
 void
 hba::handle_interrupt()
 {
+	uint32_t status = m_data->int_status;
 	for (auto &port : m_ports) {
-		if (m_data->int_status & (1 << port.index())) {
+		if (status & (1 << port.index())) {
 			port.handle_interrupt();
 		}
 	}
-	m_data->int_status = 0;
+	// Write 1 to the handled interrupts
+	m_data->int_status = status;
+}
+
+void
+hba::dump()
+{
+	console *cons = console::get();
+	static const char *regs[] = {
+		" CAP", " GHC", "  IS", "  PI", "  VS", " C3C",
+		" C3P", " EML", " EMC", "CAP2", "BOHC"
+	};
+
+	cons->printf("HBA Registers:\n");
+	auto *data = reinterpret_cast<uint32_t volatile *>(m_data);
+	for (int i = 0; i < 11; ++i) {
+		cons->printf("  %s: %08x\n", regs[i], data[i]);
+	}
+	cons->putc('\n');
 }
 
 } // namespace ahci

src/drivers/ahci/hba.h

@@ -25,12 +25,12 @@ public:
 	/// Interrupt handler.
 	void handle_interrupt();
 
-	/// Debug: find the first disk
-	port * find_disk();
+	/// Dump the HBA registers to the console
+	void dump();
 
 private:
 	pci_device *m_device;
-	hba_data *m_data;
+	hba_data volatile *m_data;
 	kutil::vector<port> m_ports;
 };
 

src/drivers/ahci/port.cpp

@@ -2,13 +2,21 @@
 #include "kutil/assert.h"
 #include "kutil/enum_bitfields.h"
 #include "ahci/ata.h"
+#include "ahci/hba.h"
 #include "ahci/fis.h"
 #include "ahci/port.h"
+#include "console.h"
 #include "io.h"
 #include "log.h"
 #include "page_manager.h"
 
+namespace ahci {
+	enum class cmd_list_flags : uint16_t;
+}
+
 IS_BITFIELD(ahci::port_cmd);
+IS_BITFIELD(volatile ahci::port_cmd);
+IS_BITFIELD(ahci::cmd_list_flags);
 
 namespace ahci {
 
@@ -110,10 +118,11 @@ struct port_data
 } __attribute__ ((packed));
 
 
-port::port(uint8_t index, port_data *data, bool impl) :
+port::port(hba *device, uint8_t index, port_data volatile *data, bool impl) :
 	m_index(index),
 	m_type(sata_signature::none),
 	m_state(state::unimpl),
+	m_hba(device),
 	m_data(data),
 	m_fis(nullptr),
 	m_cmd_list(nullptr),
@@ -154,7 +163,13 @@ port::update()
 
 		rebase();
 		m_pending.set_size(32);
-		m_data->interrupt_enable = 1;
+		for (auto &pend : m_pending) {
+			pend.type = command_type::none;
+		}
+
+		// Clear any old pending interrupts and enable interrupts
+		m_data->interrupt_status = m_data->interrupt_status;
+		m_data->interrupt_enable = 0xffffffff;
 	} else {
 		m_state = state::inactive;
 	}
@@ -190,19 +205,27 @@ port::stop_commands()
 }
 
 int
-port::make_command(size_t length)
+port::make_command(size_t length, fis_register_h2d **fis)
 {
 	int slot = -1;
-	uint32_t used_slots = (m_data->serial_active | m_data->cmd_issue);
+	uint32_t used_slots =
+		m_data->serial_active |
+		m_data->cmd_issue |
+		m_data->interrupt_status;
+
 	for (int i = 0; i < 32; ++i) {
-		if ((used_slots & (1 << i)) == 0) {
+		if (used_slots & (1 << i)) continue;
+
+		if (m_pending[i].type == command_type::none) {
 			slot = i;
 			break;
+		} else {
+			log::debug(logs::driver, "Type is %d", m_pending[i].type);
 		}
 	}
 
 	if (slot < 0) {
-		log::info(logs::driver, "AHCI could not get a free command slot.");
+		log::error(logs::driver, "AHCI could not get a free command slot.");
 		return -1;
 	}
 
@@ -214,6 +237,14 @@ port::make_command(size_t length)
 	kutil::memset(&cmdt, 0, sizeof(cmd_table) +
 			max_prd_count * sizeof(prdt_entry));
 
+	ent.flags = cmd_list_fis_size(sizeof(fis_register_h2d));
+
+	fis_register_h2d *cfis = reinterpret_cast<fis_register_h2d *>(&cmdt.cmd_fis);
+	kutil::memset(cfis, 0, sizeof(fis_register_h2d));
+	cfis->type = fis_type::register_h2d;
+	cfis->pm_port = 0x80; // set command register flag
+	*fis = cfis;
+
 	size_t remaining = length;
 	for (int i = 0; i < max_prd_count; ++i) {
 		size_t prd_len = std::min(remaining, 0x200000ul);
@@ -222,7 +253,7 @@ port::make_command(size_t length)
 		void *mem = pm->map_offset_pages(page_count(prd_len));
 		kutil::memset(mem, 0xaf, prd_len);
 
-		addr_t phys = pm->offset_phys(mem);
+		uintptr_t phys = pm->offset_phys(mem);
 		cmdt.entries[i].data_base_low = phys & 0xffffffff;
 		cmdt.entries[i].data_base_high = phys >> 32;
 		cmdt.entries[i].byte_count = prd_len - 1;
@@ -240,21 +271,20 @@ port::make_command(size_t length)
 	return slot;
 }
 
-bool
-port::read(uint64_t sector, size_t length, void *dest)
+int
+port::read_async(uint64_t offset, size_t length, void *dest)
 {
-	int slot = make_command(length);
+	fis_register_h2d *fis;
+	int slot = make_command(length, &fis);
 	if (slot < 0)
-		return false;
+		return 0;
 
 	cmd_table &cmdt = m_cmd_table[slot];
 
-	fis_register_h2d *fis = reinterpret_cast<fis_register_h2d *>(&cmdt.cmd_fis);
-	fis->type = fis_type::register_h2d;
-	fis->pm_port = 0x80; // set command register flag
 	fis->command = ata_cmd::read_dma_ext;
 	fis->device = 0x40; // ATA8-ACS p.175
 
+	uint64_t sector = offset >> 9;
 	fis->lba0 = (sector      ) & 0xff;
 	fis->lba1 = (sector >>  8) & 0xff;
 	fis->lba2 = (sector >> 16) & 0xff;
@@ -270,8 +300,64 @@ port::read(uint64_t sector, size_t length, void *dest)
 			count, sector, sector*512);
 
 	m_pending[slot].type = command_type::read;
+	m_pending[slot].offset = offset % 512;
+	m_pending[slot].count = 0;
 	m_pending[slot].data = dest;
-	return issue_command(slot);
+	if(issue_command(slot))
+		return slot;
+	else
+		return -1;
+}
+
+size_t
+port::read(uint64_t offset, size_t length, void *dest)
+{
+	int slot = read_async(offset, length, dest);
+
+	int timeout = 0;
+	while (m_pending[slot].type == command_type::read) {
+		if (timeout++ > 5) {
+			return 0;
+		}
+		asm("hlt");
+	}
+	kassert(m_pending[slot].type == command_type::finished,
+			"Read got unexpected command type");
+
+	size_t count = m_pending[slot].count;
+	m_pending[slot].type = command_type::none;
+	m_pending[slot].count = 0;
+
+	return count;
+}
+
+int
+port::identify_async()
+{
+	fis_register_h2d *fis;
+	int slot = make_command(512, &fis);
+	if (slot < 0)
+		return 0;
+
+	fis->command = ata_cmd::identify;
+
+	m_pending[slot].type = command_type::identify;
+	m_pending[slot].offset = 0;
+	m_pending[slot].count = 0;
+	m_pending[slot].data = 0;
+
+	if(issue_command(slot))
+		return slot;
+	else
+		return -1;
+}
+
+void
+port::sata_reconnect()
+{
+	m_data->serial_control |= 1;
+	io_wait(1000); // About 1ms
+	m_data->serial_control &= ~1;
 }
 
 bool
@@ -297,14 +383,23 @@ port::issue_command(int slot)
 void
 port::handle_interrupt()
 {
-	log::debug(logs::driver, "AHCI port %d got an interrupt");
+	log::debug(logs::driver, "AHCI port %d got an interrupt", m_index);
 
 	// TODO: handle other states in interrupt_status
 
-	if (m_data->interrupt_status & 0x40000000) {
+	uint32_t is = m_data->interrupt_status;
+
+	if (is & 0x00000040) {
+		// Port connect status change: For now clear the "exchange"
+		// bit in SERR, this should probably kick off diagnostics.
+		m_data->serial_error = 0x04000000;
+		identify_async();
+	}
+
+	if (is & 0x40000000) {
 		log::error(logs::driver, "AHCI task file error");
-		// TODO: clean up!
-		return;
+		dump();
+		kassert(0, "Task file error");
 	}
 
 	log::debug(logs::driver, "AHCI interrupt status: %08lx  %08lx",
@@ -312,19 +407,25 @@ port::handle_interrupt()
 
 	uint32_t ci = m_data->cmd_issue;
 	for (int i = 0; i < 32; ++i) {
+		// Skip commands still listed as "issued"
 		if (ci & (1 << i)) continue;
 
+		// Any commands not still listed as "issued" that are still pending for
+		// the driver are now finished, so handle them.
 		pending &p = m_pending[i];
 		switch (p.type) {
 		case command_type::read:
 			finish_read(i);
 			break;
+		case command_type::identify:
+			finish_identify(i);
+			break;
 		default:
 			break;
 		}
-		p.type = command_type::none;
-		p.data = nullptr;
 	}
+
+	// Clear the whole status register to mark it as handled
 	m_data->interrupt_status = m_data->interrupt_status;
 }
 
@@ -335,21 +436,92 @@ port::finish_read(int slot)
 	cmd_table &cmdt = m_cmd_table[slot];
 	cmd_list_entry &ent = m_cmd_list[slot];
 
+	size_t count = 0;
 	void *p = m_pending[slot].data;
+	uint8_t offset = m_pending[slot].offset;
 	for (int i = 0; i < ent.prd_table_length; ++i) {
 		size_t prd_len = (cmdt.entries[i].byte_count & 0x7fffffff) + 1;
 
-		addr_t phys = 
-			static_cast<addr_t>(cmdt.entries[i].data_base_low) |
-			static_cast<addr_t>(cmdt.entries[i].data_base_high) << 32;
+		uintptr_t phys =
+			static_cast<uintptr_t>(cmdt.entries[i].data_base_low) |
+			static_cast<uintptr_t>(cmdt.entries[i].data_base_high) << 32;
+
+		void *mem = kutil::offset_pointer(pm->offset_virt(phys), offset);
+
+		log::debug(logs::driver, "Reading PRD %2d: %016lx->%016lx [%lxb]", i, mem, p, prd_len);
 
-		void *mem = pm->offset_virt(phys);
 		kutil::memcpy(p, mem, prd_len);
-		p = kutil::offset_pointer(p, prd_len);
+		p = kutil::offset_pointer(p, prd_len - offset);
+		count += (prd_len - offset);
+		offset = 0;
 
 		pm->unmap_pages(mem, page_count(prd_len));
 	}
 
+	m_pending[slot].count = count;
+	m_pending[slot].type = command_type::finished;
+	m_pending[slot].data = nullptr;
+}
+
+static void
+ident_strcpy(uint16_t *from, int words, char *dest)
+{
+	for (int i = 0; i < words; ++i) {
+		*dest++ = *from >> 8;
+		*dest++ = *from & 0xff;
+		from++;
+	}
+	*dest = 0;
+}
+
+void
+port::finish_identify(int slot)
+{
+	page_manager *pm = page_manager::get();
+	cmd_table &cmdt = m_cmd_table[slot];
+	cmd_list_entry &ent = m_cmd_list[slot];
+
+	kassert(ent.prd_table_length == 1, "AHCI identify used multiple PRDs");
+
+	size_t prd_len = (cmdt.entries[0].byte_count & 0x7fffffff) + 1;
+
+	uintptr_t phys =
+		static_cast<uintptr_t>(cmdt.entries[0].data_base_low) |
+		static_cast<uintptr_t>(cmdt.entries[0].data_base_high) << 32;
+
+	log::debug(logs::driver, "Reading ident PRD:");
+
+	uint16_t *mem = reinterpret_cast<uint16_t *>(pm->offset_virt(phys));
+	char string[41];
+
+	ident_strcpy(&mem[10], 10, &string[0]);
+	log::debug(logs::driver, "    Device serial: %s", string);
+
+	ident_strcpy(&mem[23], 4, &string[0]);
+	log::debug(logs::driver, "   Device version: %s", string);
+
+	ident_strcpy(&mem[27], 20, &string[0]);
+	log::debug(logs::driver, "     Device model: %s", string);
+
+	uint32_t sectors = mem[60] | (mem[61] << 16);
+	log::debug(logs::driver, "      Max sectors: %xh", sectors);
+
+	uint16_t lb_size = mem[106];
+	log::debug(logs::driver, " lsects per psect: %d %s %s", 1 << (lb_size & 0xf),
+			lb_size & 0x20 ? "multiple logical per physical" : "",
+			lb_size & 0x10 ? "physical > 512b" : "");
+
+	uint32_t b_per_ls = 2 * (mem[117] | (mem[118] << 16));
+	log::debug(logs::driver, "      b per lsect: %d", b_per_ls);
+
+	/*
+	for (int i=0; i<256; i += 4)
+		log::debug(logs::driver, "  %3d: %04x  %3d: %04x  %3d: %04x  %3d: %04x",
+				i, mem[i], i+1, mem[i+1], i+2, mem[i+2], i+3, mem[i+3]);
+	*/
+
+	pm->unmap_pages(mem, page_count(prd_len));
+	m_pending[slot].type = command_type::none;
 }
 
 void
@@ -363,9 +535,9 @@ port::free_command(int slot)
 	for (int i = 0; i < ent.prd_table_length; ++i) {
 		size_t prd_len = (cmdt.entries[i].byte_count & 0x7fffffff) + 1;
 
-		addr_t phys =
-			static_cast<addr_t>(cmdt.entries[i].data_base_low) |
-			static_cast<addr_t>(cmdt.entries[i].data_base_high) << 32;
+		uintptr_t phys =
+			static_cast<uintptr_t>(cmdt.entries[i].data_base_low) |
+			static_cast<uintptr_t>(cmdt.entries[i].data_base_high) << 32;
 		void *mem = pm->offset_virt(phys);
 		pm->unmap_pages(mem, page_count(prd_len));
 	}
@@ -385,7 +557,7 @@ port::rebase()
 	size_t pages = 1 + page_count(prd_size * 32);
 
 	void *mem = pm->map_offset_pages(pages);
-	addr_t phys = pm->offset_phys(mem);
+	uintptr_t phys = pm->offset_phys(mem);
 
 	log::debug(logs::driver, "Rebasing address for AHCI port %d to %lx [%d]", m_index, mem, pages);
 
@@ -430,4 +602,22 @@ port::rebase()
 	start_commands();
 }
 
+void
+port::dump()
+{
+	console *cons = console::get();
+	static const char *regs[] = {
+		" CLB", "+CLB", "  FB", " +FB", "  IS", "  IE",
+		" CMD", nullptr, " TFD", " SIG", "SSTS", "SCTL", "SERR",
+		"SACT", "  CI", "SNTF", " FBS", "DEVS"
+	};
+
+	cons->printf("Port Registers:\n");
+	auto *data = reinterpret_cast<volatile uint32_t *>(m_data);
+	for (int i = 0; i < 18; ++i) {
+		if (regs[i]) cons->printf("  %s: %08x\n", regs[i], data[i]);
+	}
+	cons->putc('\n');
+}
+
 } // namespace ahci

src/drivers/ahci/port.h

@@ -4,25 +4,30 @@
 #include <stddef.h>
 #include <stdint.h>
 #include "kutil/vector.h"
+#include "block_device.h"
 
 namespace ahci {
 
 struct cmd_list_entry;
 struct cmd_table;
+struct fis_register_h2d;
+class hba;
 enum class sata_signature : uint32_t;
 enum class port_cmd : uint32_t;
 struct port_data;
 
 
 /// A port on an AHCI HBA
-class port
+class port :
+	public block_device
 {
 public:
 	/// Constructor.
-	/// \arg index Index of the port on its HBA
-	/// \arg data  Pointer to the device's registers for this port
-	/// \arg impl  Whether this port is marked as implemented in the HBA
-	port(uint8_t index, port_data *data, bool impl);
+	/// \arg device The HBA device this port belongs to
+	/// \arg index  Index of the port on its HBA
+	/// \arg data   Pointer to the device's registers for this port
+	/// \arg impl   Whether this port is marked as implemented in the HBA
+	port(hba *device, uint8_t index, port_data volatile *data, bool impl);
 
 	/// Destructor
 	~port();
@@ -37,6 +42,10 @@ public:
 	/// \returns  An enum representing the state
 	inline state get_state() const { return m_state; }
 
+	/// Check if this device is active
+	/// \returns  True if the device state is active
+	inline bool active() const { return m_state == state::active; }
+
 	/// Get the type signature of this device
 	/// \returns  An enum representing the type of device
 	inline sata_signature get_type() const { return m_type; }
@@ -53,16 +62,34 @@ public:
 	/// Stop command processing from this port
 	void stop_commands();
 
-	/// Read data from the drive.
-	/// \arg sector  Starting sector to read
+	/// Start a read operation from the drive.
+	/// \arg offset  Offset to start from
 	/// \arg length  Number of bytes to read
 	/// \arg dest    A buffer where the data will be placed
-	/// \returns     True if the command succeeded
-	bool read(uint64_t sector, size_t length, void *dest);
+	/// \returns     A handle to the read operation, or -1 on error
+	int read_async(uint64_t offset, size_t length, void *dest);
+
+	/// Read from the drive, blocking until finished.
+	/// \arg offset  Offset to start from
+	/// \arg length  Number of bytes to read
+	/// \arg dest    A buffer where the data will be placed
+	/// \returns     The number of bytes read
+	virtual size_t read(uint64_t offset, size_t length, void *dest);
+
+	/// Start an identify operation for the drive.
+	/// \returns     A handle to the read operation, or -1 on error
+	int identify_async();
+
+	/// Tell the HBA to reconnect to the SATA device. A successful
+	/// reconnect will kick off an identify command.
+	void sata_reconnect();
 
 	/// Handle an incoming interrupt
 	void handle_interrupt();
 
+	/// Dump the port registers to the console
+	void dump();
+
 private:
 	/// Rebase the port command structures to a new location in system
 	/// memory, to be allocated from the page manager.
@@ -70,8 +97,9 @@ private:
 
 	/// Initialize a command structure
 	/// \arg length  The number of bytes of data needed in the PRDs
+	/// \arg fis     [out] The FIS for this command
 	/// \returns     The index of the command slot, or -1 if none available
-	int make_command(size_t length);
+	int make_command(size_t length, fis_register_h2d **fis);
 
 	/// Send a constructed command to the hardware
 	/// \arg slot   The index of the command slot used
@@ -88,20 +116,29 @@ private:
 	/// \arg slot  The command slot that the read command used
 	void finish_read(int slot);
 
+	/// Finish an identify command started by `identify_async()`.
+	/// This will free the structures allocated, so `free_command()` is
+	/// not necessary.
+	/// \arg slot  The command slot that the read command used
+	void finish_identify(int slot);
+
 	sata_signature m_type;
 	uint8_t m_index;
 	state m_state;
 
-	port_data *m_data;
+	hba *m_hba;
+	port_data volatile *m_data;
 	void *m_fis;
 	cmd_list_entry *m_cmd_list;
 	cmd_table *m_cmd_table;
 
-	enum class command_type : uint8_t { none, read, write };
+	enum class command_type : uint8_t { none, read, write, identify, finished };
 
 	struct pending
 	{
 		command_type type;
+		uint8_t offset;
+		size_t count;
 		void *data;
 	};
 

src/drivers/nulldrv/main.cpp

@@ -0,0 +1,24 @@
+#include <stdint.h>
+#include <stdlib.h>
+
+extern "C" {
+	int32_t getpid();
+	int32_t fork();
+	void sleep(uint64_t til);
+	void debug();
+	void message(const char *msg);
+
+	int main(int, const char **);
+}
+
+
+int
+main(int argc, const char **argv)
+{
+	int32_t pid = getpid();
+	int32_t child = fork();
+	message("hello from nulldrv!");
+	for (int i = 1; i < 5; ++i)
+		sleep(i*10);
+	return 0;
+}

src/drivers/nulldrv/main.s

@@ -0,0 +1,79 @@
+section .bss
+mymessage:
+	resq 1024
+
+extern main
+extern exit
+
+section .text
+global getpid
+getpid:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x02  ; getpid syscall
+	syscall        ; pid is now already in rax, so just return
+
+	pop rbp
+	ret
+
+global debug
+debug:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x00  ; debug syscall
+	syscall
+
+	pop rbp
+	ret
+
+global sleep
+sleep:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x21  ; sleep syscall
+	syscall
+
+	pop rbp
+	ret
+
+global fork
+fork:
+	push rbp
+	mov rbp, rsp
+
+	mov rax, 0x03
+	syscall    ; pid left in rax
+
+	pop rbp
+	ret
+	
+
+global message
+message:
+	push rbp
+	mov rbp, rsp
+
+	; message should already be in rdi
+	mov rax, 0x10
+	syscall
+
+	pop rbp
+	ret
+
+
+global _start
+_start:
+	xor rbp, rbp	; Sentinel rbp
+	push rbp
+	push rbp
+	mov rbp, rsp
+
+	mov rdi, 0
+	mov rsi, 0
+	call main
+
+	mov rdi, rax
+	call exit

src/include/efi/efidef.h

@@ -20,9 +20,9 @@ Revision History
 
 --*/
 
-typedef UINT16          CHAR16;
-typedef UINT8           CHAR8;
-typedef UINT8           BOOLEAN;
+typedef wchar_t         CHAR16;
+typedef char            CHAR8;
+typedef uint8_t         BOOLEAN;
 #ifndef CONST
    #define CONST const
 #endif

src/include/efi/efisetjmp.h

@@ -4,7 +4,16 @@
 #include <efi/eficompiler.h>
 #include <efi/efisetjmp_arch.h>
 
+#ifndef __has_builtin
+#define __has_builtin(x) 0
+#endif
+
+#if ! __has_builtin(setjmp)
 extern UINTN setjmp(jmp_buf *env) __attribute__((returns_twice));
+#endif
+
+#if ! __has_builtin(longjmp)
 extern VOID longjmp(jmp_buf *env, UINTN value) __attribute__((noreturn));
+#endif
 
 #endif /* GNU_EFI_SETJMP_H */

src/include/elf.h

@@ -0,0 +1,48 @@
+#pragma once
+
+/* elf.h - basic defines for external code written assuming <elf.h> works. Only
+ * Elf64 values are included.
+ */
+
+typedef uint16_t Elf64_Half;
+typedef uint32_t Elf64_Word;
+typedef int32_t Elf64_Sword;
+typedef uint64_t Elf64_Xword;
+typedef int64_t Elf64_Sxword;
+typedef uint64_t Elf64_Addr;
+typedef uint64_t Elf64_Off;
+typedef uint16_t Elf64_Section;
+typedef Elf64_Half Elf64_Versym;
+
+typedef struct {
+	Elf64_Addr r_offset;
+	Elf64_Xword r_info;
+} Elf64_Rel;
+
+typedef struct {
+	Elf64_Addr r_offset;
+	Elf64_Word r_info;
+	Elf64_Sword r_addend;
+} Elf64_Rela;
+
+typedef struct {
+	Elf64_Sxword d_tag;
+	union {
+		Elf64_Xword d_val;
+		Elf64_Addr d_ptr;
+	} d_un;
+} Elf64_Dyn;
+
+#define ELF64_R_TYPE(x) ((x) & 0xffffffff)
+
+typedef enum {
+	DT_NULL = 0,
+	DT_RELA = 7,
+	DT_RELASZ = 8,
+	DT_RELAENT = 9
+} ElfDynTag;
+
+typedef enum {
+	R_X86_64_NONE = 0,
+	R_X86_64_RELATIVE = 8
+} Elf_x86_64_RelType;

src/include/kernel_args.h

@@ -7,17 +7,20 @@
 #define DATA_HEADER_MAGIC     0x600dda7a
 #define DATA_HEADER_VERSION   1
 
+#define POPCORN_FLAG_DEBUG    0x00000001
+
 #pragma pack(push, 1)
-struct popcorn_data {
+struct kernel_args {
 	uint32_t magic;
 	uint16_t version;
 	uint16_t length;
 
-	uint32_t _reserved0;
+	uint16_t _reserved0;
+	uint16_t scratch_pages;
 	uint32_t flags;
 
-	void *font;
-	size_t font_length;
+	void *initrd;
+	size_t initrd_length;
 
 	void *data;
 	size_t data_length;

src/include/kernel_memory.h

@@ -0,0 +1,32 @@
+#pragma once
+/// \file kernel_memory.h
+/// Constants related to the kernel's memory layout
+
+#include <stddef.h>
+#include <stdint.h>
+
+namespace memory {
+
+	/// Size of a single page frame.
+	static const size_t frame_size = 0x1000;
+
+	/// Start of kernel memory.
+	static const uintptr_t kernel_offset = 0xffffff0000000000;
+
+	/// Offset from physical where page tables are mapped.
+	static const uintptr_t page_offset = 0xffffff8000000000;
+
+	/// Initial process thread's stack address
+	static const uintptr_t initial_stack = 0x0000800000000000;
+
+	/// Initial process thread's stack size, in pages
+	static const unsigned initial_stack_pages = 1;
+
+	/// Max size of the kernel heap
+	static const size_t kernel_max_heap = 0x800000000; // 32GiB
+
+	/// Helper to determine if a physical address can be accessed
+	/// through the page_offset area.
+	inline bool page_mappable(uintptr_t a) { return (a & page_offset) == 0; }
+
+} // namespace memory

src/include/log_areas.inc

@@ -0,0 +1,9 @@
+LOG(apic,   info);
+LOG(device, info);
+LOG(paging, warn);
+LOG(driver, info);
+LOG(memory, info);
+LOG(fs,     info);
+LOG(task,   info);
+LOG(boot,   info);
+LOG(syscall,info);

src/kernel/allocator.cpp

@@ -1,9 +0,0 @@
-#include "kutil/memory_manager.h"
-
-kutil::memory_manager g_kernel_memory_manager;
-
-// kutil malloc/free implementation
-namespace kutil {
-	void * malloc(size_t n) { return g_kernel_memory_manager.allocate(n); }
-	void free(void *p) { g_kernel_memory_manager.free(p); }
-}

src/kernel/apic.cpp

@@ -1,6 +1,7 @@
 #include "kutil/assert.h"
 #include "apic.h"
 #include "interrupts.h"
+#include "io.h"
 #include "log.h"
 #include "page_manager.h"
 
@@ -45,35 +46,102 @@ apic::apic(uint32_t *base) :
 lapic::lapic(uint32_t *base, isr spurious) :
 	apic(base)
 {
+	// TODO: This causes a "reserved" page fault under KVM
 	apic_write(m_base, 0xf0, static_cast<uint32_t>(spurious));
 	log::info(logs::apic, "LAPIC created, base %lx", m_base);
 }
 
 void
-lapic::enable_timer(isr vector, uint8_t divisor, uint32_t count, bool repeat)
+lapic::calibrate_timer()
 {
+	interrupts_disable();
+
+	log::info(logs::apic, "Calibrating APIC timer...");
+
+	// Set up PIT sleep
+	uint8_t command = 0x30; // channel 0, loybyte/highbyte, mode 0
+	outb(0x43, command);
+
+	const uint32_t initial = -1u;
+	enable_timer_internal(isr::isrSpurious, 1, initial, false);
+
+	const int iterations = 5;
+	for (int i=0; i<iterations; ++i) {
+		const uint16_t pit_33ms = 39375;
+		uint16_t pit_count = pit_33ms;
+		outb(0x40, pit_count & 0xff);
+		io_wait();
+		outb(0x40, (pit_count >> 8) & 0xff);
+
+
+		while (pit_count <= pit_33ms) {
+			outb(0x43, 0); // latch counter values
+			pit_count =
+				static_cast<uint16_t>(inb(0x40)) |
+				static_cast<uint16_t>(inb(0x40)) << 8;
+		}
+	}
+
+	uint32_t remain = stop_timer();
+	uint32_t ticks_total = initial - remain;
+	m_ticks_per_us = ticks_total / (iterations * 33000);
+	log::info(logs::apic, "APIC timer ticks %d times per nanosecond.", m_ticks_per_us);
+
+	interrupts_enable();
+}
+
+uint32_t
+lapic::enable_timer_internal(isr vector, uint8_t divisor, uint32_t count, bool repeat)
+{
+	uint32_t divbits = 0;
+
 	switch (divisor) {
-	case   1: divisor = 11; break;
-	case   2: divisor = 0; break;
-	case   4: divisor = 1; break;
-	case   8: divisor = 2; break;
-	case  16: divisor = 3; break;
-	case  32: divisor = 8; break;
-	case  64: divisor = 9; break;
-	case 128: divisor = 10; break;
+	case   1: divbits = 0xb; break;
+	case   2: divbits = 0x0; break;
+	case   4: divbits = 0x1; break;
+	case   8: divbits = 0x2; break;
+	case  16: divbits = 0x3; break;
+	case  32: divbits = 0x8; break;
+	case  64: divbits = 0x9; break;
+	case 128: divbits = 0xa; break;
 	default:
 		kassert(0, "Invalid divisor passed to lapic::enable_timer");
 	}
 
-	apic_write(m_base, 0x3e0, divisor);
-	apic_write(m_base, 0x380, count);
-
 	uint32_t lvte = static_cast<uint8_t>(vector);
 	if (repeat)
 		lvte |= 0x20000;
 
-	log::debug(logs::apic, "Enabling APIC timer with isr %d.", vector);
+	log::debug(logs::apic, "Enabling APIC timer count %ld, divisor %d, isr %02x",
+			count, divisor, vector);
+
 	apic_write(m_base, 0x320, lvte);
+	apic_write(m_base, 0x3e0, divbits);
+
+	reset_timer(count);
+	return count;
+}
+
+uint32_t
+lapic::enable_timer(isr vector, uint64_t interval, bool repeat)
+{
+	uint64_t ticks = interval * m_ticks_per_us;
+
+	int divisor = 1;
+	while (ticks > -1u) {
+		ticks /= 2;
+		divisor *= 2;
+	}
+
+	return enable_timer_internal(vector, divisor, static_cast<uint32_t>(ticks), repeat);
+}
+
+uint32_t
+lapic::reset_timer(uint32_t count)
+{
+	uint32_t remaining = apic_read(m_base, 0x390);
+	apic_write(m_base, 0x380, count);
+	return remaining;
 }
 
 void

src/kernel/apic.h

@@ -32,10 +32,19 @@ public:
 
 	/// Enable interrupts for the LAPIC timer.
 	/// \arg vector   Interrupt vector the timer should use
-	/// \arg divisor  The frequency divisor of the bus Hz (power of 2, <= 128)
-	/// \arg count    The count of ticks before an interrupt
+	/// \arg interval The timer interval, in microseconds
 	/// \arg repeat   If false, this timer is one-off, otherwise repeating
-	void enable_timer(isr vector, uint8_t divisor, uint32_t count, bool repeat = true);
+	/// \returns      The count of ticks the timer is set for
+	uint32_t enable_timer(isr vector, uint64_t interval, bool repeat = true);
+
+	/// Reset the timer countdown.
+	/// \arg count  The count of ticks before an interrupt, or 0 to stop the timer
+	/// \returns    The count of ticks that were remaining before reset
+	uint32_t reset_timer(uint32_t count);
+
+	/// Stop the timer.
+	/// \returns  The count of ticks remaining before an interrupt was to happen
+	inline uint32_t stop_timer() { return reset_timer(0); }
 
 	/// Enable interrupts for the LAPIC LINT0 pin.
 	/// \arg num      Local interrupt number (0 or 1)
@@ -46,6 +55,14 @@ public:
 
 	void enable();  ///< Enable servicing of interrupts
 	void disable(); ///< Disable (temporarily) servicing of interrupts
+
+	/// Calibrate the timer speed against the PIT
+	void calibrate_timer();
+
+private:
+	uint32_t enable_timer_internal(isr vector, uint8_t divisor, uint32_t count, bool repeat);
+
+	uint32_t m_ticks_per_us;
 };
 
 

src/kernel/assert.cpp

@@ -7,23 +7,15 @@ __kernel_assert(const char *file, unsigned line, const char *message)
 	console *cons = console::get();
 	if (cons) {
 		cons->set_color(9 , 0);
-		cons->puts("\n\n  ERROR: ");
+		cons->puts("\n\n       ERROR: ");
+		cons->puts(message);
+		cons->puts("\n              ");
 		cons->puts(file);
 		cons->puts(":");
 		cons->put_dec(line);
-		cons->puts(":  ");
-		cons->puts(message);
+		cons->puts("\n");
 	}
 
-	__asm__ __volatile__( 
-		"movq %0, %%r8;"
-		"movq %1, %%r9;"
-		"movq %2, %%r10;"
-		"movq $0, %%rdx;"
-		"divq %%rdx;"
-		: // no outputs
-		: "r"((uint64_t)line), "r"(file), "r"(message)
-		: "rax", "rdx", "r8", "r9", "r10");
-
-	while (1);
+	while (1) __asm__ ("hlt");
+	__asm__ ( "int $0xe4" );
 }

src/kernel/block_device.h

@@ -0,0 +1,11 @@
+#pragma once
+/// \file block_device.h
+/// Interface definition for block devices
+#include <stddef.h>
+
+/// Interface for block devices
+class block_device
+{
+public:
+	virtual size_t read(size_t offset, size_t length, void *buffer) = 0;
+};

src/kernel/boot.s

@@ -18,18 +18,18 @@ global _start:function (_start.end - _start)
 _start:
 	cli
 
-	mov rsp, stack_end
-	push 0 ; signal end of stack with 0 return address
-	push 0 ; and a few extra entries in case of stack
-	push 0 ; problems
-	push 0
+	mov rsp, idle_stack_end
+	mov qword [rsp + 0x00], 0 ; signal end of stack with 0 return address
+	mov qword [rsp + 0x08], 0 ; and a few extra entries in case of stack
+	mov qword [rsp + 0x10], 0 ; problems
+	mov qword [rsp + 0x18], 0
 
 	mov rbp, rsp
 	extern kernel_main
 	call kernel_main
 
-	cli
-
+	; Kernel init is over, wait for the scheduler to
+	; take over
 .hang:
 	hlt
 	jmp .hang
@@ -47,6 +47,9 @@ interrupts_disable:
 
 section .bss
 align 0x100
-stack_begin:
-	resb 0x4000 ; 16KiB stack space
-stack_end:
+idle_stack_begin:
+	resb 0x1000 ; 4KiB stack space
+
+global idle_stack_end
+idle_stack_end:
+	resq 4

src/kernel/console.cpp

@@ -1,8 +1,9 @@
 #include "kutil/coord.h"
+#include "kutil/guid.h"
 #include "kutil/memory.h"
+#include "kutil/printf.h"
 #include "console.h"
 #include "font.h"
-#include "memory.h"
 #include "screen.h"
 #include "serial.h"
 
@@ -278,83 +279,9 @@ console::putc(char c)
 void console::vprintf(const char *fmt, va_list args)
 {
 	static const unsigned buf_size = 256;
-	char buffer[256];
-
-	const char *r = fmt;
-	char *w = buffer;
-	char *wend = buffer + buf_size;
-
-#define flush() do { *w = 0; puts(buffer); w = buffer; } while(0)
-
-	while (r && *r) {
-		if (w == wend) flush();
-
-		if (*r != '%') {
-			*w++ = *r++;
-			continue;
-		}
-
-		r++; // chomp the %
-		flush();
-
-		bool done = false;
-		bool right = true;
-		int width = 0;
-		char pad = ' ';
-
-		while (!done) {
-			char c = *r++;
-			switch (c) {
-				case '%': *w = '%'; done = true; break;
-
-				case '0':
-					if (width == 0) pad = '0';
-					// else fall through
-				case '1':
-				case '2':
-				case '3':
-				case '4':
-				case '5':
-				case '6':
-				case '7':
-				case '8':
-				case '9': width = width * 10 + (c - '0'); break;
-
-				case '-': right = !right; break;
-
-				case 'x': put_hex<uint32_t>(va_arg(args, uint32_t), right ? width : -width, pad); done = true; break;
-
-				case 'd':
-				case 'u': put_dec<uint32_t>(va_arg(args, uint32_t), right ? width : -width, pad); done = true; break;
-
-				case 's': {
-						const char *s = va_arg(args, const char*);
-						if (s) puts(s);
-					}
-					done = true;
-					break;
-
-				case 'l':
-					switch (*r++) {
-						case 'x': put_hex<uint64_t>(va_arg(args, uint64_t), right ? width : -width, pad); done = true; break;
-
-						case 'd':
-						case 'u': put_dec<uint32_t>(va_arg(args, uint64_t), right ? width : -width, pad); done = true; break;
-
-						default:
-							done = true;
-							break;
-					}
-					break;
-
-				default:
-					done = true;
-					break;
-			}
-		}
-	}
-
-	flush();
+	char buffer[buf_size];
+	vsnprintf_(buffer, buf_size, fmt, args);
+	puts(buffer);
 }
 
 void

src/kernel/cpu.cpp

@@ -3,6 +3,8 @@
 #include "cpu.h"
 #include "log.h"
 
+cpu_data bsp_cpu_data;
+
 inline static void
 __cpuid(
 	uint32_t leaf,
@@ -52,6 +54,7 @@ cpu_id::cpu_id()
 
 }
 
+
 cpu_id::regs
 cpu_id::get(uint32_t leaf, uint32_t sub) const
 {

src/kernel/cpu.h

@@ -1,5 +1,27 @@
 #pragma once
 
+#include <stdint.h>
+
+struct process;
+
+struct cpu_state
+{
+	uint64_t r15, r14, r13, r12, r11, r10, r9,  r8;
+	uint64_t rdi, rsi, rbp, rbx, rdx, rcx, rax;
+	uint64_t interrupt, errorcode;
+	uint64_t rip, cs, rflags, user_rsp, ss;
+};
+
+/// Per-cpu state data. If you change this, remember to update the assembly
+/// version in 'tasking.inc'
+struct cpu_data
+{
+	uintptr_t rsp0;
+	uintptr_t rsp3;
+	process *tcb;
+};
+
+extern cpu_data bsp_cpu_data;
 
 class cpu_id
 {

src/kernel/crti.s

@@ -0,0 +1,14 @@
+section .init
+global _init:function
+_init:
+	push rbp
+	mov rbp, rsp
+	; Control flow falls through to other .init sections
+
+section .fini
+global _fini:function
+_fini:
+	push rbp
+	mov rbp, rsp
+	; Control flow falls through to other .fini sections
+

src/kernel/crtn.s

@@ -0,0 +1,10 @@
+section .init
+	; Control flow falls through to here from other .init sections
+	pop rbp
+	ret
+
+section .fini
+	; Control flow falls through to here from other .fini sections
+	pop rbp
+	ret
+

src/kernel/debug.cpp

@@ -0,0 +1,81 @@
+#include "console.h"
+#include "cpu.h"
+#include "debug.h"
+#include "gdt.h"
+#include "page_manager.h"
+
+size_t __counter_syscall_enter = 0;
+size_t __counter_syscall_sysret = 0;
+
+void
+print_regs(const cpu_state &regs)
+{
+	console *cons = console::get();
+
+	uint64_t cr2 = 0;
+	__asm__ __volatile__ ("mov %%cr2, %0" : "=r"(cr2));
+
+	print_regL("rax", regs.rax);
+	print_regM("rbx", regs.rbx);
+	print_regR("rcx", regs.rcx);
+	print_regL("rdx", regs.rdx);
+	print_regM("rdi", regs.rdi);
+	print_regR("rsi", regs.rsi);
+
+	cons->puts("\n");
+	print_regL(" r8", regs.r8);
+	print_regM(" r9", regs.r9);
+	print_regR("r10", regs.r10);
+	print_regL("r11", regs.r11);
+	print_regM("r12", regs.r12);
+	print_regR("r13", regs.r13);
+	print_regL("r14", regs.r14);
+	print_regM("r15", regs.r15);
+
+	cons->puts("\n\n");
+	print_regL("rbp", regs.rbp);
+	print_regM("rsp", regs.user_rsp);
+	print_regR("sp0", bsp_cpu_data.rsp0);
+
+	print_regL("rip", regs.rip);
+	print_regM("cr3", page_manager::get()->get_pml4());
+	print_regR("cr2", cr2);
+
+	cons->puts("\n");
+}
+
+struct frame
+{
+	frame *prev;
+	uintptr_t return_addr;
+};
+
+void
+print_stacktrace(int skip)
+{
+	console *cons = console::get();
+
+	frame *fp = nullptr;
+	int fi = -skip;
+	__asm__ __volatile__ ( "mov %%rbp, %0" : "=r" (fp) );
+
+	while (fp && fp->return_addr) {
+		if (fi++ >= 0)
+			cons->printf("  frame %2d: %lx\n", fi-1, fp->return_addr);
+		fp = fp->prev;
+	}
+}
+
+void
+print_stack(const cpu_state &regs)
+{
+	console *cons = console::get();
+
+	cons->puts("\nStack:\n");
+	uint64_t sp = regs.user_rsp;
+	while (sp <= regs.rbp) {
+		cons->printf("%016x: %016x\n", sp, *reinterpret_cast<uint64_t *>(sp));
+		sp += sizeof(uint64_t);
+	}
+}
+

src/kernel/debug.h

@@ -0,0 +1,25 @@
+#pragma once
+/// \file debug.h
+/// Debugging utilities
+
+#include <stdint.h>
+
+extern "C" {
+	uintptr_t get_rsp();
+	uintptr_t get_rip();
+	uintptr_t get_frame(int frame);
+	uintptr_t get_gsbase();
+	void _halt();
+}
+
+extern size_t __counter_syscall_enter;
+extern size_t __counter_syscall_sysret;
+
+void print_regs(const cpu_state &regs);
+void print_stack(const cpu_state &regs);
+void print_stacktrace(int skip);
+
+#define print_regL(name, value) cons->printf("       %s: %016lx", name, (value));
+#define print_regM(name, value) cons->printf("    %s: %016lx", name, (value));
+#define print_regR(name, value) cons->printf("    %s: %016lx\n", name, (value));
+

src/kernel/debug.s

@@ -1,29 +1,20 @@
+global get_rsp
+get_rsp:
+	mov rax, rsp
+	ret
 
-section .text
-global do_the_set_registers
-do_the_set_registers:
-	mov rax, 0xdeadbeef0badc0de
-	mov r8, rcx
-	mov r9, rdi
+global get_rip
+get_rip:
+	pop rax ; do the same thing as 'ret', except with 'jmp'
+	jmp rax ; with the return address still in rax
+
+global get_gsbase
+get_gsbase:
+	rdgsbase rax
+	ret
 
 global _halt
 _halt:
 	hlt
 	jmp _halt
 
-
-global get_frame
-get_frame:
-	mov rcx, rbp
-
-.loop:
-	mov rax, [rcx + 8]
-	mov rcx, [rcx]
-	cmp rdi, 0
-	je .done
-
-	sub rdi, 1
-	jmp .loop
-
-.done:
-	ret

src/kernel/device_manager.cpp

@@ -4,20 +4,17 @@
 #include "kutil/assert.h"
 #include "kutil/memory.h"
 #include "acpi_tables.h"
-#include "ahci/driver.h"
 #include "apic.h"
 #include "console.h"
 #include "device_manager.h"
 #include "interrupts.h"
 #include "log.h"
-#include "memory.h"
 #include "page_manager.h"
 
 
 static const char expected_signature[] = "RSD PTR ";
 
-device_manager device_manager::s_instance(nullptr);
-ahci_driver ahcid;
+device_manager device_manager::s_instance(nullptr, kutil::allocator::invalid);
 
 struct acpi1_rsdp
 {
@@ -42,29 +39,16 @@ struct acpi2_rsdp
 	uint8_t reserved[3];
 } __attribute__ ((packed));
 
-uint8_t
-acpi_checksum(const void *p, size_t len, size_t off = 0)
-{
-	uint8_t sum = 0;
-	const uint8_t *c = reinterpret_cast<const uint8_t *>(p);
-	for (int i = off; i < len; ++i) sum += c[i];
-	return sum;
-}
-
 bool
 acpi_table_header::validate(uint32_t expected_type) const
 {
-	if (acpi_checksum(this, length) != 0) return false;
+	if (kutil::checksum(this, length) != 0) return false;
 	return !expected_type || (expected_type == type);
 }
 
 
 void irq2_callback(void *)
 {
-	console *cons = console::get();
-	cons->set_color(11);
-	cons->puts(".");
-	cons->set_color();
 }
 
 void irq4_callback(void *)
@@ -75,8 +59,13 @@ void irq4_callback(void *)
 }
 
 
-device_manager::device_manager(const void *root_table) :
-	m_lapic(nullptr)
+device_manager::device_manager(const void *root_table, kutil::allocator &alloc) :
+	m_lapic(nullptr),
+	m_ioapics(alloc),
+	m_pci(alloc),
+	m_devices(alloc),
+	m_irqs(alloc),
+	m_blockdevs(alloc)
 {
 	kassert(root_table != 0, "ACPI root table pointer is null.");
 
@@ -87,7 +76,7 @@ device_manager::device_manager(const void *root_table) :
 		kassert(acpi1->signature[i] == expected_signature[i],
 				"ACPI RSDP table signature mismatch");
 
-	uint8_t sum = acpi_checksum(acpi1, sizeof(acpi1_rsdp), 0);
+	uint8_t sum = kutil::checksum(acpi1, sizeof(acpi1_rsdp), 0);
 	kassert(sum == 0, "ACPI 1.0 RSDP checksum mismatch.");
 
 	kassert(acpi1->revision > 1, "ACPI 1.0 not supported.");
@@ -95,7 +84,7 @@ device_manager::device_manager(const void *root_table) :
 	const acpi2_rsdp *acpi2 =
 		reinterpret_cast<const acpi2_rsdp *>(acpi1);
 
-	sum = acpi_checksum(acpi2, sizeof(acpi2_rsdp), sizeof(acpi1_rsdp));
+	sum = kutil::checksum(acpi2, sizeof(acpi2_rsdp), sizeof(acpi1_rsdp));
 	kassert(sum == 0, "ACPI 2.0 RSDP checksum mismatch.");
 
 	load_xsdt(reinterpret_cast<const acpi_xsdt *>(acpi2->xsdt_address));
@@ -109,7 +98,7 @@ device_manager::device_manager(const void *root_table) :
 ioapic *
 device_manager::get_ioapic(int i)
 {
-	return (i < m_ioapics.count()) ? m_ioapics[i] : nullptr;
+	return (i < m_ioapics.count()) ? &m_ioapics[i] : nullptr;
 }
 
 static void
@@ -164,26 +153,44 @@ device_manager::load_apic(const acpi_apic *apic)
 	uint8_t const *p = apic->controller_data;
 	uint8_t const *end = p + count;
 
-	// Pass one: set up IOAPIC objcts
+	// Pass one: count IOAPIC objcts
+	int num_ioapics = 0;
+	while (p < end) {
+		const uint8_t type = p[0];
+		const uint8_t length = p[1];
+		if (type == 1) num_ioapics++;
+		p += length;
+	}
+
+	m_ioapics.set_capacity(num_ioapics);
+
+	// Pass two: set up IOAPIC objcts
+	p = apic->controller_data;
 	while (p < end) {
 		const uint8_t type = p[0];
 		const uint8_t length = p[1];
 		if (type == 1) {
 			uint32_t *base = reinterpret_cast<uint32_t *>(kutil::read_from<uint32_t>(p+4));
 			uint32_t base_gsr = kutil::read_from<uint32_t>(p+8);
-			m_ioapics.append(new ioapic(base, base_gsr));
+			m_ioapics.emplace(base, base_gsr);
 		}
 		p += length;
 	}
 
-	// Pass two: configure APIC objects
+	// Pass three: configure APIC objects
 	p = apic->controller_data;
 	while (p < end) {
 		const uint8_t type = p[0];
 		const uint8_t length = p[1];
 
 		switch (type) {
-		case 0: // Local APIC
+		case 0: { // Local APIC
+				uint8_t uid = kutil::read_from<uint8_t>(p+2);
+				uint8_t id = kutil::read_from<uint8_t>(p+3);
+				log::debug(logs::device, "    Local APIC uid %x id %x", id);
+			}
+			break;
+
 		case 1: // I/O APIC
 			break;
 
@@ -196,7 +203,7 @@ device_manager::load_apic(const acpi_apic *apic)
 						source, gsi, (flags & 0x3), ((flags >> 2) & 0x3));
 
 				// TODO: in a multiple-IOAPIC system this might be elsewhere
-				m_ioapics[0]->redirect(source, static_cast<isr>(gsi), flags, true);
+				m_ioapics[0].redirect(source, static_cast<isr>(gsi), flags, true);
 			}
 			break;
 
@@ -222,16 +229,13 @@ device_manager::load_apic(const acpi_apic *apic)
 		p += length;
 	}
 
-	// m_lapic->enable_timer(isr::isrTimer, 128, 3000000);
-
-	for (uint8_t i = 0; i < m_ioapics[0]->get_num_gsi(); ++i) {
+	for (uint8_t i = 0; i < m_ioapics[0].get_num_gsi(); ++i) {
 		switch (i) {
 			case 2: break;
-			default: m_ioapics[0]->mask(i, false);
+			default: m_ioapics[0].mask(i, false);
 		}
 	}
 
-	m_ioapics[0]->dump_redirs();
 	m_lapic->enable();
 }
 
@@ -292,6 +296,7 @@ device_manager::init_drivers()
 {
 	// Eventually this should be e.g. a lookup into a loadable driver list
 	// for now, just look for AHCI devices
+	/*
 	for (auto &device : m_devices) {
 		if (device.devclass() != 1 || device.subclass() != 6)
 			continue;
@@ -303,6 +308,7 @@ device_manager::init_drivers()
 
 		ahcid.register_device(&device);
 	}
+	*/
 }
 
 bool
@@ -320,3 +326,9 @@ device_manager::allocate_msi(const char *name, pci_device &device, irq_callback
 			static_cast<uint16_t>(vector));
 	return true;
 }
+
+void
+device_manager::register_block_device(block_device *blockdev)
+{
+	m_blockdevs.append(blockdev);
+}

src/kernel/device_manager.h

@@ -2,13 +2,13 @@
 /// \file device_manager.h
 /// The device manager definition
 #include "kutil/vector.h"
+#include "apic.h"
 #include "pci.h"
 
 struct acpi_xsdt;
 struct acpi_apic;
 struct acpi_mcfg;
-class lapic;
-class ioapic;
+class block_device;
 
 using irq_callback = void (*)(void *);
 
@@ -18,8 +18,9 @@ class device_manager
 {
 public:
 	/// Constructor.
-	/// \arg root_table   Pointer to the ACPI RSDP
-	device_manager(const void *root_table);
+	/// \arg root_table  Pointer to the ACPI RSDP
+	/// \arg alloc       Allocator for device arrays
+	device_manager(const void *root_table, kutil::allocator &alloc);
 
 	/// Get the system global device manager.
 	/// \returns  A reference to the system device manager
@@ -65,6 +66,23 @@ public:
 		return false;
 	}
 
+	/// Register the existance of a block device.
+	/// \arg blockdev  Pointer to the block device
+	void register_block_device(block_device *blockdev);
+
+	/// Get the number of block devices in the system
+	/// \returns  A count of devices
+	inline unsigned get_num_block_devices() const { return m_blockdevs.count(); }
+
+	/// Get a block device
+	/// \arg i    Index of the device to get
+	/// \returns  A pointer to the requested device, or nullptr
+	inline block_device * get_block_device(unsigned i)
+	{
+		return i < m_blockdevs.count() ?
+			m_blockdevs[i] : nullptr;
+	}
+
 private:
 	/// Parse the ACPI XSDT and load relevant sub-tables.
 	/// \arg xsdt  Pointer to the XSDT from the firmware
@@ -87,7 +105,7 @@ private:
 	void bad_irq(uint8_t irq);
 
 	lapic *m_lapic;
-	kutil::vector<ioapic *> m_ioapics;
+	kutil::vector<ioapic> m_ioapics;
 
 	kutil::vector<pci_group> m_pci;
 	kutil::vector<pci_device> m_devices;
@@ -100,6 +118,8 @@ private:
 	};
 	kutil::vector<irq_allocation> m_irqs;
 
+	kutil::vector<block_device *> m_blockdevs;
+
 	static device_manager s_instance;
 
 	device_manager() = delete;

src/kernel/frame_allocator.cpp

@@ -0,0 +1,93 @@
+#include "kutil/assert.h"
+#include "kutil/memory.h"
+#include "frame_allocator.h"
+
+using memory::frame_size;
+using memory::page_offset;
+using frame_block_node = kutil::list_node<frame_block>;
+
+frame_allocator g_frame_allocator;
+
+int
+frame_block::compare(const frame_block *rhs) const
+{
+	if (address < rhs->address)
+		return -1;
+	else if (address > rhs->address)
+		return 1;
+	return 0;
+}
+
+
+frame_allocator::raw_alloc::raw_alloc(frame_allocator &fa) : m_fa(fa) {}
+
+void *
+frame_allocator::raw_alloc::allocate(size_t size)
+{
+	kassert(size <= frame_size, "Raw allocator only allocates a single page");
+
+	uintptr_t addr = 0;
+	if (size <= frame_size)
+		m_fa.allocate(1, &addr);
+	return reinterpret_cast<void*>(addr + page_offset);
+}
+
+void
+frame_allocator::raw_alloc::free(void *p)
+{
+	m_fa.free(reinterpret_cast<uintptr_t>(p), 1);
+}
+
+
+frame_allocator::frame_allocator() :
+	m_raw_alloc(*this)
+{
+}
+
+size_t
+frame_allocator::allocate(size_t count, uintptr_t *address)
+{
+	kassert(!m_free.empty(), "frame_allocator::pop_frames ran out of free frames!");
+	if (m_free.empty())
+		return 0;
+
+	auto *first = m_free.front();
+
+	if (count >= first->count) {
+		*address = first->address;
+		m_free.remove(first);
+		return first->count;
+	} else {
+		first->count -= count;
+		*address = first->address + (first->count * frame_size);
+		return count;
+	}
+}
+
+inline uintptr_t end(frame_block *node) { return node->address + node->count * frame_size; }
+
+void
+frame_allocator::free(uintptr_t address, size_t count)
+{
+	frame_block_node *node =
+		reinterpret_cast<frame_block_node*>(address + page_offset);
+
+	kutil::memset(node, 0, sizeof(frame_block_node));
+	node->address = address;
+	node->count = count;
+
+	m_free.sorted_insert(node);
+
+	frame_block_node *next = node->next();
+	if (next && end(node) == next->address) {
+		node->count += next->count;
+		m_free.remove(next);
+	}
+
+	frame_block_node *prev = node->prev();
+	if (prev && end(prev) == address) {
+		prev->count += node->count;
+		m_free.remove(node);
+	}
+}
+

src/kernel/frame_allocator.h

@@ -0,0 +1,70 @@
+#pragma once
+/// \file frame_allocator.h
+/// Allocator for physical memory frames
+
+#include <stdint.h>
+
+#include "kutil/allocator.h"
+#include "kutil/linked_list.h"
+
+struct frame_block;
+using frame_block_list = kutil::linked_list<frame_block>;
+
+/// Allocator for physical memory frames
+class frame_allocator
+{
+public:
+	/// Default constructor
+	frame_allocator();
+
+	/// Get free frames from the free list. Only frames from the first free block
+	/// are returned, so the number may be less than requested, but they will
+	/// be contiguous.
+	/// \arg count    The maximum number of frames to get
+	/// \arg address  [out] The physical address of the first frame
+	/// \returns      The number of frames retrieved
+	size_t allocate(size_t count, uintptr_t *address);
+
+	/// Free previously allocated frames.
+	/// \arg address  The physical address of the first frame to free
+	/// \arg count    The number of frames to be freed
+	void free(uintptr_t address, size_t count);
+
+	/// Get a memory allocator that allocates raw pages
+	/// \returns  The allocator ojbect
+	kutil::allocator & raw_allocator() { return m_raw_alloc; }
+
+private:
+	class raw_alloc :
+		public kutil::allocator
+	{
+	public:
+		raw_alloc(frame_allocator &fa);
+		virtual void * allocate(size_t size) override;
+		virtual void free(void *p) override;
+	private:
+		frame_allocator &m_fa;
+	};
+
+	raw_alloc m_raw_alloc;
+	frame_block_list m_free; ///< Free frames list
+
+	frame_allocator(const frame_allocator &) = delete;
+};
+
+
+/// A block of contiguous frames. Each `frame_block` represents contiguous
+/// physical frames with the same attributes.
+struct frame_block
+{
+	uintptr_t address;
+	uint32_t count;
+
+	/// Compare two blocks by address.
+	/// \arg rhs   The right-hand comparator
+	/// \returns   <0 if this is sorts earlier, >0 if this sorts later, 0 for equal
+	int compare(const frame_block *rhs) const;
+};
+
+
+extern frame_allocator g_frame_allocator;

src/kernel/fs/gpt.cpp

@@ -0,0 +1,114 @@
+#include "kutil/assert.h"
+#include "kutil/guid.h"
+#include "kutil/memory.h"
+#include "device_manager.h"
+#include "fs/gpt.h"
+#include "log.h"
+
+namespace fs {
+
+const kutil::guid efi_system_part = kutil::make_guid(0xC12A7328, 0xF81F, 0x11D2, 0xBA4B, 0x00A0C93EC93B);
+const kutil::guid efi_unused_part = kutil::make_guid(0, 0, 0, 0, 0);
+
+const uint64_t gpt_signature = 0x5452415020494645; // "EFI PART"
+const size_t block_size = 512;
+
+struct gpt_header
+{
+	uint64_t signature;
+	uint32_t revision;
+	uint32_t headersize;
+	uint32_t crc32;
+	uint32_t reserved;
+	uint64_t my_lba;
+	uint64_t alt_lba;
+	uint64_t first_usable_lba;
+	uint64_t last_usable_lba;
+	kutil::guid disk_guid;
+	uint64_t table_lba;
+	uint32_t entry_count;
+	uint32_t entry_length;
+	uint32_t table_crc32;
+} __attribute__ ((packed));
+
+struct gpt_entry
+{
+	kutil::guid type;
+	kutil::guid part_guid;
+	uint64_t start_lba;
+	uint64_t end_lba;
+	uint64_t attributes;
+	uint16_t name_wide[36];
+} __attribute__ ((packed));
+
+
+partition::partition(block_device *parent, size_t start, size_t length) :
+	m_parent(parent),
+	m_start(start),
+	m_length(length)
+{
+}
+
+size_t
+partition::read(size_t offset, size_t length, void *buffer)
+{
+	if (offset + length > m_length)
+		offset = m_length - offset;
+	return m_parent->read(m_start + offset, length, buffer);
+}
+
+
+unsigned
+partition::load(block_device *device)
+{
+	// Read LBA 1
+	uint8_t block[block_size];
+	size_t count = device->read(block_size, block_size, &block);
+	kassert(count == block_size, "Short read for GPT header.");
+
+	gpt_header *header = reinterpret_cast<gpt_header *>(&block);
+	if (header->signature != gpt_signature)
+		return 0;
+
+	size_t arraysize = header->entry_length * header->entry_count;
+	log::debug(logs::fs, "Found GPT header: %d paritions, size 0x%lx",
+			header->entry_count, arraysize);
+
+	uint8_t *array = new uint8_t[arraysize];
+	count = device->read(block_size * header->table_lba, arraysize, array);
+	kassert(count == arraysize, "Short read for GPT entry array.");
+
+	auto &dm = device_manager::get();
+
+	unsigned found = 0;
+	gpt_entry *entry0 = reinterpret_cast<gpt_entry *>(array);
+	for (uint32_t i = 0; i < header->entry_count; ++i) {
+		gpt_entry *entry = kutil::offset_pointer(entry0, i * header->entry_length);
+		if (entry->type == efi_unused_part) continue;
+
+		// TODO: real UTF16->UTF8
+		char name[sizeof(gpt_entry::name_wide) / 2];
+		for (int i = 0; i < sizeof(name); ++i)
+			name[i] = entry->name_wide[i];
+
+		log::debug(logs::fs, "Found partition %02x at %lx-%lx", i, entry->start_lba, entry->end_lba);
+		if (entry->type == efi_system_part)
+			log::debug(logs::fs, "   type EFI SYSTEM PARTITION");
+		else
+			log::debug(logs::fs, "   type %G", entry->type);
+		log::debug(logs::fs, "   name %s", name);
+		log::debug(logs::fs, "   attr %016lx", entry->attributes);
+
+		found += 1;
+		partition *part = new partition(
+				device,
+				entry->start_lba * block_size,
+				(entry->end_lba - entry->start_lba) * block_size);
+		dm.register_block_device(part);
+	}
+
+	return found;
+}
+
+
+} // namespace fs

src/kernel/fs/gpt.h

@@ -0,0 +1,37 @@
+#pragma once
+/// \file gpt.h
+/// Definitions for dealing with GUID Partition Tables
+#include "block_device.h"
+
+namespace fs {
+
+
+class partition :
+	public block_device
+{
+public:
+	/// Constructor.
+	/// \arg parent  The block device this partition is a part of
+	/// \arg start   The starting offset in bytes from the start of the parent
+	/// \arg lenght  The length in bytes of this partition
+	partition(block_device *parent, size_t start, size_t length);
+
+	/// Read bytes from the partition.
+	/// \arg offset  The offset in bytes at which to start reading
+	/// \arg length  The number of bytes to read
+	/// \arg buffer  [out] Data is read into this buffer
+	/// \returns     The number of bytes read
+	virtual size_t read(size_t offset, size_t length, void *buffer);
+
+	/// Find partitions on a block device and add them to the device manager
+	/// \arg device  The device to search for partitions
+	/// \returns     The number of partitions found
+	static unsigned load(block_device *device);
+
+private:
+	block_device *m_parent;
+	size_t m_start;
+	size_t m_length;
+};
+
+} // namespace fs

src/kernel/gdt.cpp

@@ -0,0 +1,288 @@
+#include <stdint.h>
+
+#include "kutil/assert.h"
+#include "kutil/enum_bitfields.h"
+#include "kutil/memory.h"
+#include "console.h"
+#include "log.h"
+
+
+enum class gdt_type : uint8_t
+{
+	accessed	= 0x01,
+	read_write	= 0x02,
+	conforming	= 0x04,
+	execute		= 0x08,
+	system		= 0x10,
+	ring1		= 0x20,
+	ring2		= 0x40,
+	ring3		= 0x60,
+	present		= 0x80
+};
+IS_BITFIELD(gdt_type);
+
+struct gdt_descriptor
+{
+	uint16_t limit_low;
+	uint16_t base_low;
+	uint8_t base_mid;
+	gdt_type type;
+	uint8_t size;
+	uint8_t base_high;
+} __attribute__ ((packed));
+
+struct tss_descriptor
+{
+	uint16_t limit_low;
+	uint16_t base_00;
+	uint8_t base_16;
+	gdt_type type;
+	uint8_t size;
+	uint8_t base_24;
+	uint32_t base_32;
+	uint32_t reserved;
+} __attribute__ ((packed));
+
+struct tss_entry
+{
+	uint32_t reserved0;
+
+	uint64_t rsp[3]; // stack pointers for CPL 0-2
+	uint64_t ist[8]; // ist[0] is reserved
+
+	uint64_t reserved1;
+	uint16_t reserved2;
+	uint16_t iomap_offset;
+} __attribute__ ((packed));
+
+struct idt_descriptor
+{
+	uint16_t base_low;
+	uint16_t selector;
+	uint8_t ist;
+	uint8_t flags;
+	uint16_t base_mid;
+	uint32_t base_high;
+	uint32_t reserved;		// must be zero
+} __attribute__ ((packed));
+
+struct table_ptr
+{
+	uint16_t limit;
+	uint64_t base;
+} __attribute__ ((packed));
+
+
+gdt_descriptor g_gdt_table[10];
+idt_descriptor g_idt_table[256];
+table_ptr g_gdtr;
+table_ptr g_idtr;
+tss_entry g_tss;
+
+
+extern "C" {
+	void idt_write();
+	void idt_load();
+
+	void gdt_write(uint16_t cs, uint16_t ds, uint16_t tr);
+	void gdt_load();
+}
+
+void
+gdt_set_entry(uint8_t i, uint32_t base, uint64_t limit, bool is64, gdt_type type)
+{
+	g_gdt_table[i].limit_low = limit & 0xffff;
+	g_gdt_table[i].size = (limit >> 16) & 0xf;
+	g_gdt_table[i].size |= (is64 ? 0xa0 : 0xc0);
+
+	g_gdt_table[i].base_low = base & 0xffff;
+	g_gdt_table[i].base_mid = (base >> 16) & 0xff;
+	g_gdt_table[i].base_high = (base >> 24) & 0xff;
+
+	g_gdt_table[i].type = type | gdt_type::system | gdt_type::present;
+}
+
+void
+tss_set_entry(uint8_t i, uint64_t base, uint64_t limit)
+{
+	tss_descriptor tssd;
+	tssd.limit_low = limit & 0xffff;
+	tssd.size = (limit >> 16) & 0xf;
+
+	tssd.base_00 = base & 0xffff;
+	tssd.base_16 = (base >> 16) & 0xff;
+	tssd.base_24 = (base >> 24) & 0xff;
+	tssd.base_32 = (base >> 32) & 0xffffffff;
+	tssd.reserved = 0;
+
+	tssd.type =
+		gdt_type::accessed |
+		gdt_type::execute |
+		gdt_type::ring3 |
+		gdt_type::present;
+	kutil::memcpy(&g_gdt_table[i], &tssd, sizeof(tss_descriptor));
+}
+
+void
+idt_set_entry(uint8_t i, uint64_t addr, uint16_t selector, uint8_t flags)
+{
+	g_idt_table[i].base_low = addr & 0xffff;
+	g_idt_table[i].base_mid = (addr >> 16) & 0xffff;
+	g_idt_table[i].base_high = (addr >> 32) & 0xffffffff;
+	g_idt_table[i].selector = selector;
+	g_idt_table[i].flags = flags;
+	g_idt_table[i].ist = 0;
+	g_idt_table[i].reserved = 0;
+}
+
+void
+tss_set_stack(int ring, uintptr_t rsp)
+{
+	kassert(ring < 3, "Bad ring passed to tss_set_stack.");
+	g_tss.rsp[ring] = rsp;
+}
+
+uintptr_t
+tss_get_stack(int ring)
+{
+	kassert(ring < 3, "Bad ring passed to tss_get_stack.");
+	return g_tss.rsp[ring];
+}
+
+void
+gdt_init()
+{
+	kutil::memset(&g_gdt_table, 0, sizeof(g_gdt_table));
+	kutil::memset(&g_idt_table, 0, sizeof(g_idt_table));
+
+	g_gdtr.limit = sizeof(g_gdt_table) - 1;
+	g_gdtr.base = reinterpret_cast<uint64_t>(&g_gdt_table);
+
+	// Kernel CS/SS - always 64bit
+	gdt_set_entry(1, 0, 0xfffff,  true, gdt_type::read_write | gdt_type::execute);
+	gdt_set_entry(2, 0, 0xfffff,  true, gdt_type::read_write);
+
+	// User CS32/SS/CS64 - layout expected by SYSRET
+	gdt_set_entry(3, 0, 0xfffff,  false, gdt_type::ring3 | gdt_type::read_write | gdt_type::execute);
+	gdt_set_entry(4, 0, 0xfffff,  true, gdt_type::ring3 | gdt_type::read_write);
+	gdt_set_entry(5, 0, 0xfffff,  true, gdt_type::ring3 | gdt_type::read_write | gdt_type::execute);
+
+	kutil::memset(&g_tss, 0, sizeof(tss_entry));
+	g_tss.iomap_offset = sizeof(tss_entry);
+
+	uintptr_t tss_base = reinterpret_cast<uintptr_t>(&g_tss);
+
+	// Note that this takes TWO GDT entries
+	tss_set_entry(6, tss_base, sizeof(tss_entry));
+
+	gdt_write(1 << 3, 2 << 3, 6 << 3);
+
+	g_idtr.limit = sizeof(g_idt_table) - 1;
+	g_idtr.base = reinterpret_cast<uint64_t>(&g_idt_table);
+
+	idt_write();
+}
+
+void
+gdt_dump(int index)
+{
+	const table_ptr &table = g_gdtr;
+
+	console *cons = console::get();
+
+	int start = 0;
+	int count = (table.limit + 1) / sizeof(gdt_descriptor);
+	if (index != -1) {
+		start = index;
+		count = 1;
+	} else {
+		cons->printf("         GDT: loc:%lx size:%d\n", table.base, table.limit+1);
+	}
+
+	const gdt_descriptor *gdt =
+		reinterpret_cast<const gdt_descriptor *>(table.base);
+
+	for (int i = start; i < start+count; ++i) {
+		uint32_t base =
+			(gdt[i].base_high << 24) |
+			(gdt[i].base_mid << 16) |
+			gdt[i].base_low;
+
+		uint32_t limit =
+			static_cast<uint32_t>(gdt[i].size & 0x0f) << 16 |
+			gdt[i].limit_low;
+
+		cons->printf("          %02d:", i);
+		if (! bitfield_has(gdt[i].type, gdt_type::present)) {
+			cons->puts(" Not Present\n");
+			continue;
+		}
+
+		cons->printf(" Base %08x  limit %05x ", base, limit);
+
+		switch (gdt[i].type & gdt_type::ring3) {
+			case gdt_type::ring3: cons->puts("ring3"); break;
+			case gdt_type::ring2: cons->puts("ring2"); break;
+			case gdt_type::ring1: cons->puts("ring1"); break;
+			default: cons->puts("ring0"); break;
+		}
+
+		cons->printf(" %s %s %s %s %s %s %s\n",
+			bitfield_has(gdt[i].type, gdt_type::accessed) ? "A" : " ",
+			bitfield_has(gdt[i].type, gdt_type::read_write) ? "RW" : "  ",
+			bitfield_has(gdt[i].type, gdt_type::conforming) ? "C" : " ",
+			bitfield_has(gdt[i].type, gdt_type::execute) ? "EX" : "  ",
+			bitfield_has(gdt[i].type, gdt_type::system) ? "S" : " ",
+			(gdt[i].size & 0x80) ? "KB" : " B",
+			(gdt[i].size & 0x60) == 0x20 ? "64" :
+				(gdt[i].size & 0x60) == 0x40 ? "32" : "16");
+	}
+}
+
+void
+idt_dump(int index)
+{
+	const table_ptr &table = g_idtr;
+
+
+	int start = 0;
+	int count = (table.limit + 1) / sizeof(idt_descriptor);
+	if (index != -1) {
+		start = index;
+		count = 1;
+		log::info(logs::boot, "IDT FOR INDEX %02x", index);
+	} else {
+		log::info(logs::boot, "Loaded IDT at: %lx size: %d bytes", table.base, table.limit+1);
+	}
+
+	const idt_descriptor *idt =
+		reinterpret_cast<const idt_descriptor *>(table.base);
+
+	for (int i = start; i < start+count; ++i) {
+		uint64_t base =
+			(static_cast<uint64_t>(idt[i].base_high) << 32) |
+			(static_cast<uint64_t>(idt[i].base_mid)  << 16) |
+			idt[i].base_low;
+
+		char const *type;
+		switch (idt[i].flags & 0xf) {
+			case 0x5: type = " 32tsk "; break;
+			case 0x6: type = " 16int "; break;
+			case 0x7: type = " 16trp "; break;
+			case 0xe: type = " 32int "; break;
+			case 0xf: type = " 32trp "; break;
+			default:  type = " ????? "; break;
+		}
+
+		if (idt[i].flags & 0x80) {
+			log::debug(logs::boot,
+					"   Entry %3d: Base:%lx Sel(rpl %d, ti %d, %3d) IST:%d %s DPL:%d", i, base,
+					(idt[i].selector & 0x3),
+					((idt[i].selector & 0x4) >> 2),
+					(idt[i].selector >> 3),
+					idt[i].ist,
+					type,
+					((idt[i].flags >> 5) & 0x3));
+		}
+	}
+}

src/kernel/gdt.h

@@ -0,0 +1,33 @@
+#pragma once
+/// \file gdt.h
+/// Definitions relating to system descriptor tables: GDT, IDT, TSS
+#include <stdint.h>
+
+/// Set up the GDT and TSS, and switch segment registers to point
+/// to them.
+void gdt_init();
+
+/// Set an entry in the IDT
+/// \arg i         Index in the IDT (vector of the interrupt this handles)
+/// \arg addr      Address of the handler
+/// \arg selector  GDT selector to set when invoking this handler
+/// \arg flags     Descriptor flags to set
+void idt_set_entry(uint8_t i, uint64_t addr, uint16_t selector, uint8_t flags);
+
+/// Set the stack pointer for a given ring in the TSS
+/// \arg ring  Ring to set for (0-2)
+/// \arg rsp   Stack pointer to set
+void tss_set_stack(int ring, uintptr_t rsp);
+
+/// Get the stack pointer for a given ring in the TSS
+/// \arg ring  Ring to get (0-2)
+/// \returns   Stack pointers for that ring
+uintptr_t tss_get_stack(int ring);
+
+/// Dump information about the current GDT to the screen
+/// \arg index  Which entry to print, or -1 for all entries
+void gdt_dump(int index = -1);
+
+/// Dump information about the current IDT to the screen
+/// \arg index  Which entry to print, or -1 for all entries
+void idt_dump(int index = -1);

src/kernel/gdt.s

@@ -0,0 +1,35 @@
+extern g_idtr
+extern g_gdtr
+
+global idt_write
+idt_write:
+	lidt [rel g_idtr]
+	ret
+
+global idt_load
+idt_load:
+	sidt [rel g_idtr]
+	ret
+
+global gdt_write
+gdt_write:
+	lgdt [rel g_gdtr]
+	mov ax, si ; second arg is data segment
+	mov ds, ax
+	mov es, ax
+	mov fs, ax
+	mov gs, ax
+	mov ss, ax
+	push qword rdi ; first arg is code segment
+	lea rax, [rel .next]
+	push rax
+	o64 retf
+.next:
+	ltr dx ; third arg is the TSS
+	ret
+
+global gdt_load
+gdt_load:
+	sgdt [rel g_gdtr]
+	ret
+

src/kernel/interrupt_isrs.inc

@@ -134,9 +134,115 @@ IRQ (0x7d, 0x5d, irq5D)
 IRQ (0x7e, 0x5e, irq5E)
 IRQ (0x7f, 0x5f, irq5F)
 
-ISR (0xec, isrTimer)
-ISR (0xed, isrLINT0)
-ISR (0xee, isrLINT1)
+IRQ (0x80, 0x60, irq60)
+IRQ (0x81, 0x61, irq61)
+IRQ (0x82, 0x62, irq62)
+IRQ (0x83, 0x63, irq63)
+IRQ (0x84, 0x64, irq64)
+IRQ (0x85, 0x65, irq65)
+IRQ (0x86, 0x66, irq66)
+IRQ (0x87, 0x67, irq67)
+IRQ (0x88, 0x68, irq68)
+IRQ (0x89, 0x69, irq69)
+IRQ (0x8a, 0x6a, irq6A)
+IRQ (0x8b, 0x6b, irq6B)
+IRQ (0x8c, 0x6c, irq6C)
+IRQ (0x8d, 0x6d, irq6D)
+IRQ (0x8e, 0x6e, irq6E)
+IRQ (0x8f, 0x6f, irq6F)
+
+IRQ (0x90, 0x70, irq70)
+IRQ (0x91, 0x71, irq71)
+IRQ (0x92, 0x72, irq72)
+IRQ (0x93, 0x73, irq73)
+IRQ (0x94, 0x74, irq74)
+IRQ (0x95, 0x75, irq75)
+IRQ (0x96, 0x76, irq76)
+IRQ (0x97, 0x77, irq77)
+IRQ (0x98, 0x78, irq78)
+IRQ (0x99, 0x79, irq79)
+IRQ (0x9a, 0x7a, irq7A)
+IRQ (0x9b, 0x7b, irq7B)
+IRQ (0x9c, 0x7c, irq7C)
+IRQ (0x9d, 0x7d, irq7D)
+IRQ (0x9e, 0x7e, irq7E)
+IRQ (0x9f, 0x7f, irq7F)
+
+IRQ (0xa0, 0x80, irq80)
+IRQ (0xa1, 0x81, irq81)
+IRQ (0xa2, 0x82, irq82)
+IRQ (0xa3, 0x83, irq83)
+IRQ (0xa4, 0x84, irq84)
+IRQ (0xa5, 0x85, irq85)
+IRQ (0xa6, 0x86, irq86)
+IRQ (0xa7, 0x87, irq87)
+IRQ (0xa8, 0x88, irq88)
+IRQ (0xa9, 0x89, irq89)
+IRQ (0xaa, 0x8a, irq8A)
+IRQ (0xab, 0x8b, irq8B)
+IRQ (0xac, 0x8c, irq8C)
+IRQ (0xad, 0x8d, irq8D)
+IRQ (0xae, 0x8e, irq8E)
+IRQ (0xaf, 0x8f, irq8F)
+
+IRQ (0xb0, 0x90, irq90)
+IRQ (0xb1, 0x91, irq91)
+IRQ (0xb2, 0x92, irq92)
+IRQ (0xb3, 0x93, irq93)
+IRQ (0xb4, 0x94, irq94)
+IRQ (0xb5, 0x95, irq95)
+IRQ (0xb6, 0x96, irq96)
+IRQ (0xb7, 0x97, irq97)
+IRQ (0xb8, 0x98, irq98)
+IRQ (0xb9, 0x99, irq99)
+IRQ (0xba, 0x9a, irq9A)
+IRQ (0xbb, 0x9b, irq9B)
+IRQ (0xbc, 0x9c, irq9C)
+IRQ (0xbd, 0x9d, irq9D)
+IRQ (0xbe, 0x9e, irq9E)
+IRQ (0xbf, 0x9f, irq9F)
+
+IRQ (0xc0, 0xa0, irqA0)
+IRQ (0xc1, 0xa1, irqA1)
+IRQ (0xc2, 0xa2, irqA2)
+IRQ (0xc3, 0xa3, irqA3)
+IRQ (0xc4, 0xa4, irqA4)
+IRQ (0xc5, 0xa5, irqA5)
+IRQ (0xc6, 0xa6, irqA6)
+IRQ (0xc7, 0xa7, irqA7)
+IRQ (0xc8, 0xa8, irqA8)
+IRQ (0xc9, 0xa9, irqA9)
+IRQ (0xca, 0xaa, irqAA)
+IRQ (0xcb, 0xab, irqAB)
+IRQ (0xcc, 0xac, irqAC)
+IRQ (0xcd, 0xad, irqAD)
+IRQ (0xce, 0xae, irqAE)
+IRQ (0xcf, 0xaf, irqAF)
+
+IRQ (0xd0, 0xb0, irqB0)
+IRQ (0xd1, 0xb1, irqB1)
+IRQ (0xd2, 0xb2, irqB2)
+IRQ (0xd3, 0xb3, irqB3)
+IRQ (0xd4, 0xb4, irqB4)
+IRQ (0xd5, 0xb5, irqB5)
+IRQ (0xd6, 0xb6, irqB6)
+IRQ (0xd7, 0xb7, irqB7)
+IRQ (0xd8, 0xb8, irqB8)
+IRQ (0xd9, 0xb9, irqB9)
+IRQ (0xda, 0xba, irqBA)
+IRQ (0xdb, 0xbb, irqBB)
+IRQ (0xdc, 0xbc, irqBC)
+IRQ (0xdd, 0xbd, irqBD)
+IRQ (0xde, 0xbe, irqBE)
+IRQ (0xdf, 0xbf, irqBF)
+
+
+ISR (0xe0, isrTimer)
+ISR (0xe1, isrLINT0)
+ISR (0xe2, isrLINT1)
+ISR (0xe4, isrAssert)
+
+UISR(0xee, isrSyscall)
 ISR (0xef, isrSpurious)
 
 ISR (0xf0, isrIgnore0)

src/kernel/interrupts.cpp

@@ -1,85 +1,37 @@
-#include <stddef.h>
 #include <stdint.h>
 
-#include "kutil/enum_bitfields.h"
-#include "kutil/memory.h"
+#include "apic.h"
 #include "console.h"
+#include "cpu.h"
+#include "debug.h"
 #include "device_manager.h"
+#include "gdt.h"
 #include "interrupts.h"
 #include "io.h"
 #include "log.h"
+#include "scheduler.h"
+#include "syscall.h"
 
-enum class gdt_flags : uint8_t
-{
-	ac = 0x01,
-	rw = 0x02,
-	dc = 0x04,
-	ex = 0x08,
-	r1 = 0x20,
-	r2 = 0x40,
-	r3 = 0x60,
-	pr = 0x80,
-
-	res_1 = 0x10
-};
-IS_BITFIELD(gdt_flags);
-
-struct gdt_descriptor
-{
-	uint16_t limit_low;
-	uint16_t base_low;
-	uint8_t base_mid;
-	uint8_t flags;
-	uint8_t granularity;
-	uint8_t base_high;
-} __attribute__ ((packed));
-
-struct idt_descriptor
-{
-	uint16_t base_low;
-	uint16_t selector;
-	uint8_t ist;
-	uint8_t flags;
-	uint16_t base_mid;
-	uint32_t base_high;
-	uint32_t reserved;		// must be zero
-} __attribute__ ((packed));
-
-struct table_ptr
-{
-	uint16_t limit;
-	uint64_t base;
-} __attribute__ ((packed));
-
-gdt_descriptor g_gdt_table[10];
-idt_descriptor g_idt_table[256];
-table_ptr g_gdtr;
-table_ptr g_idtr;
-
-struct registers;
+static const uint16_t PIC1 = 0x20;
+static const uint16_t PIC2 = 0xa0;
 
 extern "C" {
-	void idt_write();
-	void idt_load();
+	void _halt();
 
-	void gdt_write();
-	void gdt_load();
-
-	void isr_handler(registers);
-	void irq_handler(registers);
+	void isr_handler(cpu_state*);
+	void irq_handler(cpu_state*);
 
 #define ISR(i, name)     extern void name ();
 #define EISR(i, name)    extern void name ();
+#define UISR(i, name)    extern void name ();
 #define IRQ(i, q, name)  extern void name ();
 #include "interrupt_isrs.inc"
 #undef IRQ
+#undef UISR
 #undef EISR
 #undef ISR
 }
 
-void idt_dump(const table_ptr &table);
-void gdt_dump(const table_ptr &table);
-
 isr
 operator+(const isr &lhs, int rhs)
 {
@@ -93,9 +45,11 @@ get_irq(unsigned vector)
 	switch (vector) {
 #define ISR(i, name)
 #define EISR(i, name)
+#define UISR(i, name)
 #define IRQ(i, q, name) case i : return q;
 #include "interrupt_isrs.inc"
 #undef IRQ
+#undef UISR
 #undef EISR
 #undef ISR
 
@@ -103,43 +57,12 @@ get_irq(unsigned vector)
 	}
 }
 
-void
-set_gdt_entry(uint8_t i, uint32_t base, uint32_t limit, bool is64, gdt_flags flags)
-{
-	g_gdt_table[i].limit_low = limit & 0xffff;
-
-	g_gdt_table[i].base_low = base & 0xffff;
-	g_gdt_table[i].base_mid = (base >> 16) & 0xff;
-	g_gdt_table[i].base_high = (base >> 24) & 0xff;
-
-	g_gdt_table[i].granularity =
-			((limit >> 16) & 0xf) | (is64 ? 0xa0 : 0xc0);
-
-	g_gdt_table[i].flags = static_cast<uint8_t>(flags | gdt_flags::res_1 | gdt_flags::pr);
-}
-
-void
-set_idt_entry(uint8_t i, uint64_t addr, uint16_t selector, uint8_t flags)
-{
-	g_idt_table[i].base_low = addr & 0xffff;
-	g_idt_table[i].base_mid = (addr >> 16) & 0xffff;
-	g_idt_table[i].base_high = (addr >> 32) & 0xffffffff;
-	g_idt_table[i].selector = selector;
-	g_idt_table[i].flags = flags;
-	g_idt_table[i].ist = 0;
-	g_idt_table[i].reserved = 0;
-}
-
 static void
 disable_legacy_pic()
 {
-
-	static const uint16_t PIC1 = 0x20;
-	static const uint16_t PIC2 = 0xa0;
-
 	// Mask all interrupts
-	outb(0xa1, 0xff);
-	outb(0x21, 0xff);
+	outb(PIC2+1, 0xfc);
+	outb(PIC1+1, 0xff);
 
 	// Start initialization sequence
 	outb(PIC1, 0x11); io_wait();
@@ -147,7 +70,7 @@ disable_legacy_pic()
 
 	// Remap into ignore ISRs
 	outb(PIC1+1, static_cast<uint8_t>(isr::isrIgnore0)); io_wait();
-	outb(PIC2+1, static_cast<uint8_t>(isr::isrIgnore0)); io_wait();
+	outb(PIC2+1, static_cast<uint8_t>(isr::isrIgnore8)); io_wait();
 
 	// Tell PICs about each other
 	outb(PIC1+1, 0x04); io_wait();
@@ -164,72 +87,124 @@ enable_serial_interrupts()
 void
 interrupts_init()
 {
-	kutil::memset(&g_gdt_table, 0, sizeof(g_gdt_table));
-	kutil::memset(&g_idt_table, 0, sizeof(g_idt_table));
-
-	g_gdtr.limit = sizeof(g_gdt_table) - 1;
-	g_gdtr.base = reinterpret_cast<uint64_t>(&g_gdt_table);
-
-	set_gdt_entry(1, 0, 0xfffff, false, gdt_flags::rw);
-	set_gdt_entry(2, 0, 0xfffff, false, gdt_flags::rw | gdt_flags::ex | gdt_flags::dc);
-	set_gdt_entry(3, 0, 0xfffff, false, gdt_flags::rw);
-	set_gdt_entry(4, 0, 0xfffff, false, gdt_flags::rw | gdt_flags::ex);
-
-	set_gdt_entry(6, 0, 0xfffff, false, gdt_flags::rw);
-	set_gdt_entry(7, 0, 0xfffff,  true, gdt_flags::rw | gdt_flags::ex);
-
-	gdt_write();
-
-	g_idtr.limit = sizeof(g_idt_table) - 1;
-	g_idtr.base = reinterpret_cast<uint64_t>(&g_idt_table);
-
-#define ISR(i, name)     set_idt_entry(i, reinterpret_cast<uint64_t>(& name), 0x38, 0x8e);
-#define EISR(i, name)    set_idt_entry(i, reinterpret_cast<uint64_t>(& name), 0x38, 0x8e);
-#define IRQ(i, q, name)  set_idt_entry(i, reinterpret_cast<uint64_t>(& name), 0x38, 0x8e);
+#define ISR(i, name)     idt_set_entry(i, reinterpret_cast<uint64_t>(& name), 0x08, 0x8e);
+#define EISR(i, name)    idt_set_entry(i, reinterpret_cast<uint64_t>(& name), 0x08, 0x8e);
+#define UISR(i, name)    idt_set_entry(i, reinterpret_cast<uint64_t>(& name), 0x08, 0xee);
+#define IRQ(i, q, name)  idt_set_entry(i, reinterpret_cast<uint64_t>(& name), 0x08, 0x8e);
 #include "interrupt_isrs.inc"
 #undef IRQ
+#undef UISR
 #undef EISR
 #undef ISR
 
-	idt_write();
 	disable_legacy_pic();
 	enable_serial_interrupts();
 
 	log::info(logs::boot, "Interrupts enabled.");
 }
 
-struct registers
-{
-	uint64_t ds;
-	uint64_t rdi, rsi, rbp, rsp, rbx, rdx, rcx, rax;
-	uint64_t interrupt, errorcode;
-	uint64_t rip, cs, eflags, user_esp, ss;
-};
-
-#define print_reg(name, value) cons->printf("         %s: %016lx\n", name, (value));
-
-extern "C" uint64_t get_frame(int frame);
-
 void
-print_stacktrace(int skip = 0)
-{
-	console *cons = console::get();
-	int frame = 0;
-	uint64_t bp = get_frame(skip);
-	while (bp) {
-		cons->printf("    frame %2d: %lx\n", frame, bp);
-		bp = get_frame(++frame + skip);
-	}
-}
-
-void
-isr_handler(registers regs)
+isr_handler(cpu_state *regs)
 {
 	console *cons = console::get();
 
-	switch (static_cast<isr>(regs.interrupt & 0xff)) {
+	switch (static_cast<isr>(regs->interrupt & 0xff)) {
+
+	case isr::isrDebug: {
+			cons->set_color(11);
+			cons->puts("\nDebug Exception:\n");
+			cons->set_color();
+
+			uint64_t dr = 0;
+
+			__asm__ __volatile__ ("mov %%dr0, %0" : "=r"(dr));
+			print_regL("dr0", dr);
+
+			__asm__ __volatile__ ("mov %%dr1, %0" : "=r"(dr));
+			print_regM("dr1", dr);
+
+			__asm__ __volatile__ ("mov %%dr2, %0" : "=r"(dr));
+			print_regM("dr2", dr);
+
+			__asm__ __volatile__ ("mov %%dr3, %0" : "=r"(dr));
+			print_regR("dr3", dr);
+
+			__asm__ __volatile__ ("mov %%dr6, %0" : "=r"(dr));
+			print_regL("dr6", dr);
+
+			__asm__ __volatile__ ("mov %%dr7, %0" : "=r"(dr));
+			print_regR("dr7", dr);
+
+			print_regL("rip", regs->rip);
+			print_regM("rsp", regs->user_rsp);
+			print_regM("fla", regs->rflags);
+			_halt();
+		}
+		break;
+
+	case isr::isrGPFault: {
+			cons->set_color(9);
+			cons->puts("\nGeneral Protection Fault:\n");
+			cons->set_color();
+
+			cons->printf("   errorcode: %lx", regs->errorcode);
+			if (regs->errorcode & 0x01) cons->puts(" external");
+
+			int index = (regs->errorcode & 0xffff) >> 4;
+			if (index) {
+				switch ((regs->errorcode & 0x07) >> 1) {
+				case 0:
+					cons->printf(" GDT[%x]\n", index);
+					gdt_dump(index);
+					break;
+
+				case 1:
+				case 3:
+					cons->printf(" IDT[%x]\n", index);
+					idt_dump(index);
+					break;
+
+				default:
+					cons->printf(" LDT[%x]??\n", index);
+					break;
+				}
+			} else {
+				cons->putc('\n');
+			}
+			print_regs(*regs);
+			/*
+			print_stacktrace(2);
+			print_stack(*regs);
+			*/
+		}
+		_halt();
+		break;
+
+	case isr::isrPageFault: {
+			uintptr_t cr2 = 0;
+			__asm__ __volatile__ ("mov %%cr2, %0" : "=r"(cr2));
+
+			if (!page_manager::get()->fault_handler(cr2)) {
+				cons->set_color(11);
+				cons->puts("\nPage Fault:\n");
+				cons->set_color();
+
+				cons->puts("       flags:");
+				if (regs->errorcode & 0x01) cons->puts(" present");
+				if (regs->errorcode & 0x02) cons->puts(" write");
+				if (regs->errorcode & 0x04) cons->puts(" user");
+				if (regs->errorcode & 0x08) cons->puts(" reserved");
+				if (regs->errorcode & 0x10) cons->puts(" ip");
+				cons->puts("\n");
+				print_regs(*regs);
+				print_stacktrace(2);
+				_halt();
+			}
+		}
+		break;
+
 	case isr::isrTimer:
-		cons->puts("\nTICK\n");
+		scheduler::get().tick();
 		break;
 
 	case isr::isrLINT0:
@@ -240,6 +215,24 @@ isr_handler(registers regs)
 		cons->puts("\nLINT1\n");
 		break;
 
+	case isr::isrAssert: {
+			cons->set_color();
+			print_regs(*regs);
+			print_stacktrace(2);
+		}
+		_halt();
+		break;
+
+	/*
+	case isr::isrSyscall:
+		syscall_dispatch(regs);
+		break;
+	*/
+
+	case isr::isrSpurious:
+		// No EOI for the spurious interrupt
+		return;
+
 	case isr::isrIgnore0:
 	case isr::isrIgnore1:
 	case isr::isrIgnore2:
@@ -248,177 +241,52 @@ isr_handler(registers regs)
 	case isr::isrIgnore5:
 	case isr::isrIgnore6:
 	case isr::isrIgnore7:
-
-		/*
-		cons->printf("\nIGNORED PIC INTERRUPT %d\n",
-				(regs.interrupt % 0xff) - 0xf0);
-		*/
+		//cons->printf("\nIGNORED: %02x\n", regs->interrupt);
+		outb(PIC1, 0x20);
 		break;
 
-	case isr::isrPageFault: {
-			cons->set_color(11);
-			cons->puts("\nPage Fault:\n");
-			cons->set_color();
-
-			cons->puts("       flags:");
-			if (regs.errorcode & 0x01) cons->puts(" present");
-			if (regs.errorcode & 0x02) cons->puts(" write");
-			if (regs.errorcode & 0x04) cons->puts(" user");
-			if (regs.errorcode & 0x08) cons->puts(" reserved");
-			if (regs.errorcode & 0x10) cons->puts(" ip");
-			cons->puts("\n");
-
-			uint64_t cr2 = 0;
-			__asm__ __volatile__ ("mov %%cr2, %0" : "=r"(cr2));
-			print_reg("cr2", cr2);
-
-			print_reg("rip", regs.rip);
-
-			cons->puts("\n");
-			print_stacktrace(2);
-		}
-		while(1) asm("hlt");
+	case isr::isrIgnore8:
+	case isr::isrIgnore9:
+	case isr::isrIgnoreA:
+	case isr::isrIgnoreB:
+	case isr::isrIgnoreC:
+	case isr::isrIgnoreD:
+	case isr::isrIgnoreE:
+	case isr::isrIgnoreF:
+		//cons->printf("\nIGNORED: %02x\n", regs->interrupt);
+		outb(PIC1, 0x20);
+		outb(PIC2, 0x20);
 		break;
 
 	default:
 		cons->set_color(9);
-		cons->puts("\nReceived ISR interrupt:\n");
+		cons->printf("\nReceived %02x interrupt:\n",
+				(static_cast<isr>(regs->interrupt)));
+
 		cons->set_color();
+		cons->printf("         ISR: %02lx  ERR: %lx\n\n",
+				regs->interrupt, regs->errorcode);
 
-		cons->printf("         ISR: %02lx\n", regs.interrupt);
-		cons->printf("         ERR: %lx\n", regs.errorcode);
-		cons->puts("\n");
-
-		print_reg(" ds", regs.ds);
-		print_reg("rdi", regs.rdi);
-		print_reg("rsi", regs.rsi);
-		print_reg("rbp", regs.rbp);
-		print_reg("rsp", regs.rsp);
-		print_reg("rbx", regs.rbx);
-		print_reg("rdx", regs.rdx);
-		print_reg("rcx", regs.rcx);
-		print_reg("rax", regs.rax);
-		cons->puts("\n");
-
-		print_reg("rip", regs.rip);
-		print_reg(" cs", regs.cs);
-		print_reg(" ef", regs.eflags);
-		print_reg("esp", regs.user_esp);
-		print_reg(" ss", regs.ss);
-
-		cons->puts("\n");
+		print_regs(*regs);
 		print_stacktrace(2);
-		while(1) asm("hlt");
+		_halt();
 	}
-
 	*reinterpret_cast<uint32_t *>(0xffffff80fee000b0) = 0;
 }
 
 void
-irq_handler(registers regs)
+irq_handler(cpu_state *regs)
 {
 	console *cons = console::get();
-	uint8_t irq = get_irq(regs.interrupt);
+	uint8_t irq = get_irq(regs->interrupt);
 	if (! device_manager::get().dispatch_irq(irq)) {
 		cons->set_color(11);
 		cons->printf("\nReceived unknown IRQ: %d (vec %d)\n",
-				irq, regs.interrupt);
+				irq, regs->interrupt);
 		cons->set_color();
-
-		print_reg(" ds", regs.ds);
-		print_reg("rdi", regs.rdi);
-		print_reg("rsi", regs.rsi);
-		print_reg("rbp", regs.rbp);
-		print_reg("rsp", regs.rsp);
-		print_reg("rbx", regs.rbx);
-		print_reg("rdx", regs.rdx);
-		print_reg("rcx", regs.rcx);
-		print_reg("rax", regs.rax);
-		cons->puts("\n");
-
-		print_reg("rip", regs.rip);
-		print_reg(" cs", regs.cs);
-		print_reg(" ef", regs.eflags);
-		print_reg("esp", regs.user_esp);
-		print_reg(" ss", regs.ss);
-		while(1) asm("hlt");
+		print_regs(*regs);
+		_halt();
 	}
 
 	*reinterpret_cast<uint32_t *>(0xffffff80fee000b0) = 0;
 }
-
-
-void
-gdt_dump(const table_ptr &table)
-{
-	log::info(logs::boot, "Loaded GDT at: %lx size: %d bytes", table.base, table.limit+1);
-
-	int count = (table.limit + 1) / sizeof(gdt_descriptor);
-	const gdt_descriptor *gdt =
-		reinterpret_cast<const gdt_descriptor *>(table.base);
-
-	for (int i = 0; i < count; ++i) {
-		uint32_t base =
-			(gdt[i].base_high << 24) |
-			(gdt[i].base_mid << 16) |
-			gdt[i].base_low;
-
-		uint32_t limit =
-			static_cast<uint32_t>(gdt[i].granularity & 0x0f) << 16 |
-			gdt[i].limit_low;
-
-		if (gdt[i].flags & 0x80) {
-			log::debug(logs::boot,
-					"   Entry %3d: Base %x  limit %x  privs %d  flags %s%s%s%s%s%s",
-					i, base, limit, ((gdt[i].flags >> 5) & 0x03),
-
-					(gdt[i].flags & 0x80) ? "P " : "  ",
-					(gdt[i].flags & 0x08) ? "ex " : "   ",
-					(gdt[i].flags & 0x04) ? "dc " : "   ",
-					(gdt[i].flags & 0x02) ? "rw " : "   ",
-
-					(gdt[i].granularity & 0x80) ? "kb " : "b  ",
-					(gdt[i].granularity & 0x60) == 0x60 ? "64" :
-						(gdt[i].granularity & 0x60) == 0x40 ? "32" : "16"
-					);
-		}
-	}
-}
-
-void
-idt_dump(const table_ptr &table)
-{
-	log::info(logs::boot, "Loaded IDT at: %lx size: %d bytes", table.base, table.limit+1);
-
-	int count = (table.limit + 1) / sizeof(idt_descriptor);
-	const idt_descriptor *idt =
-		reinterpret_cast<const idt_descriptor *>(table.base);
-
-	for (int i = 0; i < count; ++i) {
-		uint64_t base =
-			(static_cast<uint64_t>(idt[i].base_high) << 32) |
-			(static_cast<uint64_t>(idt[i].base_mid)  << 16) |
-			idt[i].base_low;
-
-		char const *type;
-		switch (idt[i].flags & 0xf) {
-			case 0x5: type = " 32tsk "; break;
-			case 0x6: type = " 16int "; break;
-			case 0x7: type = " 16trp "; break;
-			case 0xe: type = " 32int "; break;
-			case 0xf: type = " 32trp "; break;
-			default:  type = " ????? "; break;
-		}
-
-		if (idt[i].flags & 0x80) {
-			log::debug(logs::boot,
-					"   Entry %3d: Base:%lx Sel(rpl %d, ti %d, %3d) IST:%d %s DPL:%d", i, base,
-					(idt[i].selector & 0x3),
-					((idt[i].selector & 0x4) >> 2),
-					(idt[i].selector >> 3),
-					idt[i].ist,
-					type,
-					((idt[i].flags >> 5) & 0x3));
-		}
-	}
-}

src/kernel/interrupts.h

@@ -1,6 +1,7 @@
 #pragma once
 /// \file interrupts.h
 /// Free functions and definitions related to interrupt service vectors
+#include <stdint.h>
 
 
 /// Enum of all defined ISR/IRQ vectors
@@ -8,20 +9,28 @@ enum class isr : uint8_t
 {
 #define ISR(i, name)     name = i,
 #define EISR(i, name)    name = i,
+#define UISR(i, name)    name = i,
 #define IRQ(i, q, name)  name = i,
 #include "interrupt_isrs.inc"
 #undef IRQ
+#undef UISR
 #undef EISR
 #undef ISR
 
 	_zero = 0
 };
 
+/// Helper operator to add an offset to an isr vector
 isr operator+(const isr &lhs, int rhs);
 
 extern "C" {
+	/// Set the CPU interrupt enable flag (sti)
 	void interrupts_enable();
+
+	/// Set the CPU interrupt disable flag (cli)
 	void interrupts_disable();
 }
 
+/// Fill the IDT with our ISRs, and disable the legacy
+/// PIC interrupts.
 void interrupts_init();

src/kernel/interrupts.s

@@ -1,97 +1,38 @@
-extern g_idtr
-extern g_gdtr
-
-global idt_write
-idt_write:
-	lidt [rel g_idtr]
-	ret
-
-global idt_load
-idt_load:
-	sidt [rel g_idtr]
-	ret
-
-global gdt_write
-gdt_write:
-	lgdt [rel g_gdtr]
-	ret
-
-global gdt_load
-gdt_load:
-	sgdt [rel g_gdtr]
-	ret
-
-%macro push_all_and_segments 0
-	push rax
-	push rcx
-	push rdx
-	push rbx
-	push rsp
-	push rbp
-	push rsi
-	push rdi
-
-	mov ax, ds
-	push rax
-%endmacro
-
-%macro pop_all_and_segments 0
-	pop rax
-	mov ds, ax
-	mov es, ax
-	mov fs, ax
-	mov gs, ax
-
-	pop rdi
-	pop rsi
-	pop rbp
-	pop rsp
-	pop rbx
-	pop rdx
-	pop rcx
-	pop rax
-%endmacro
-
-%macro load_kernel_segments 0
-	mov ax, 0x10	; load the kernel data segment
-	mov ds, ax
-	mov es, ax
-	mov fs, ax
-	mov gs, ax
-%endmacro
+%include "push_all.inc"
 
 extern isr_handler
 global isr_handler_prelude
 isr_handler_prelude:
-	push_all_and_segments
-	load_kernel_segments
+	push_all
+	check_swap_gs
 
+	mov rdi, rsp
+	mov rsi, rsp
 	call isr_handler
-
-	pop_all_and_segments
-
-	add rsp, 16		; because the ISRs added err/num
-	sti
-	iretq
+	jmp isr_handler_return
 
 extern irq_handler
 global irq_handler_prelude
 irq_handler_prelude:
-	push_all_and_segments
-	load_kernel_segments
+	push_all
+	check_swap_gs
 
+	mov rdi, rsp
+	mov rsi, rsp
 	call irq_handler
+	; fall through to isr_handler_return
 
-	pop_all_and_segments
+global isr_handler_return
+isr_handler_return:
+	check_swap_gs
+	pop_all
 
 	add rsp, 16		; because the ISRs added err/num
-	sti
 	iretq
 
 %macro EMIT_ISR 2
 	global %1
 	%1:
-		cli
 		push 0
 		push %2
 		jmp isr_handler_prelude
@@ -100,7 +41,6 @@ irq_handler_prelude:
 %macro EMIT_EISR 2
 	global %1
 	%1:
-		cli
 		push %2
 		jmp isr_handler_prelude
 %endmacro
@@ -108,13 +48,13 @@ irq_handler_prelude:
 %macro EMIT_IRQ 2
 	global %1
 	%1:
-		cli
 		push 0
 		push %2
 		jmp irq_handler_prelude
 %endmacro
 
-%define EISR(i, name)     EMIT_EISR name, i
+%define EISR(i, name)     EMIT_EISR name, i   ; ISR with error code
+%define UISR(i, name)     EMIT_ISR name, i    ; ISR callable from user space
 %define  ISR(i, name)     EMIT_ISR name, i
 %define  IRQ(i, q, name)  EMIT_IRQ name, i
 

src/kernel/io.cpp

@@ -14,24 +14,9 @@ outb(uint16_t port, uint8_t val)
 	__asm__ __volatile__ ( "outb %0, %1" :: "a"(val), "Nd"(port) );
 }
 
-uint64_t
-rdmsr(uint64_t addr)
-{
-	uint32_t low, high;
-	__asm__ __volatile__ ("rdmsr" : "=a"(low), "=d"(high) : "c"(addr));
-	return (static_cast<uint64_t>(high) << 32) | low;
-}
-
 void
-wrmsr(uint64_t addr, uint64_t value)
+io_wait(unsigned times)
 {
-	uint32_t low = value & 0xffffffff;
-	uint32_t high = value >> 32;
-	__asm__ __volatile__ ("wrmsr" :: "c"(addr), "a"(low), "d"(high));
-}
-
-void
-io_wait()
-{
-	outb(0x80, 0);
+	for (unsigned i = 0; i < times; ++i)
+		outb(0x80, 0);
 }

src/kernel/io.h

@@ -14,18 +14,9 @@ uint8_t inb(uint16_t port);
 /// \arg val   The byte to write
 void outb(uint16_t port, uint8_t val);
 
-/// Read the value of a MSR
-/// \arg addr   The MSR address
-/// \returns    The current value of the MSR
-uint64_t rdmsr(uint64_t addr);
-
-/// Write to a MSR
-/// \arg addr   The MSR address
-/// \arg value  The value to write
-void wrmsr(uint64_t addr, uint64_t value);
-
 /// Pause briefly by doing IO to port 0x80
-void io_wait();
+/// \arg times  Number of times to delay by writing
+void io_wait(unsigned times = 1);
 
 }
 

src/kernel/loader.s

@@ -0,0 +1,22 @@
+%include "push_all.inc"
+
+extern load_process_image
+
+global ramdisk_process_loader
+ramdisk_process_loader:
+
+	; create_process already pushed a cpu_state onto the stack for us, this
+	; acts both as the cpu_state parameter to load_process_image, and the
+	; saved state for the following iretq
+
+	pop rdi ; the address of the program image
+	pop rsi ; the size of the program image
+	pop rdx ; the address of this process' process structure
+
+	call load_process_image
+
+	push rax ; load_process_image returns the process entrypoint
+
+	swapgs
+	iretq
+

src/kernel/log.cpp

@@ -1,104 +1,55 @@
-#include <type_traits>
-#include "kutil/assert.h"
 #include "kutil/memory.h"
 #include "console.h"
 #include "log.h"
+#include "scheduler.h"
 
 
-static const uint64_t default_enabled[] = {0x00, 0xff, 0xff, 0xff};
-static const uint8_t level_colors[] = {0x07, 0x0f, 0x0b, 0x09};
+static uint8_t log_buffer[0x10000];
+static log::logger g_logger(log_buffer, sizeof(log_buffer));
 
-static const char *levels[] = {"debug", " info", " warn", "error", "fatal"};
-static const char *areas[] = {
-	"boot",
-	"mem ",
-	"apic",
-	"dev ",
-	"driv",
+static const uint8_t level_colors[] = {0x07, 0x07, 0x0f, 0x0b, 0x09};
 
-	nullptr
-};
-
-
-log log::s_log;
-
-log::log() : m_cons(nullptr)
+static void
+output_log(log::area_t area, log::level severity, const char *message)
 {
-	kassert(0, "Invalid log constructor");
-}
-
-log::log(console *cons) :
-	m_cons(cons)
-{
-	const int num_levels = static_cast<int>(level::max) - 1;
-	for (int i = 0; i < num_levels; ++i)
-		m_enabled[i] = default_enabled[i];
+	auto *cons = console::get();
+	cons->set_color(level_colors[static_cast<int>(severity)]);
+	cons->printf("%7s %5s: %s\n",
+			g_logger.area_name(area),
+			g_logger.level_name(severity),
+			message);
+	cons->set_color();
 }
 
 void
-log::init(console *cons)
+logger_task()
 {
-	new (&s_log) log(cons);
-	log::info(logs::boot, "Logging system initialized.");
-}
+	uint8_t buffer[257];
+	auto *ent = reinterpret_cast<log::logger::entry *>(buffer);
+	auto *cons = console::get();
 
-static inline uint64_t
-bit_mask(logs area) { return 1 << static_cast<uint64_t>(area); }
+	g_logger.set_immediate(nullptr);
+	log::info(logs::task, "Starting kernel logger task");
 
-void
-log::enable(logs type, level at_level)
-{
-	using under_t = std::underlying_type<level>::type;
-	under_t at = static_cast<under_t>(at_level);
-	under_t max = sizeof(m_enabled) / sizeof(m_enabled[0]);
+	scheduler &s = scheduler::get();
 
-	for (under_t i = 0; i < max; ++i) {
-		if (i >= at)
-			s_log.m_enabled[i] |= bit_mask(type);
-		else
-			s_log.m_enabled[i] &= ~bit_mask(type);
+	while (true) {
+		if(g_logger.get_entry(buffer, sizeof(buffer))) {
+			buffer[ent->bytes] = 0;
+			cons->set_color(level_colors[static_cast<int>(ent->severity)]);
+			cons->printf("%7s %5s: %s\n",
+				g_logger.area_name(ent->area),
+				g_logger.level_name(ent->severity),
+				ent->message);
+			cons->set_color();
+		} else {
+			s.schedule();
+		}
 	}
 }
 
-template <>
-void
-log::trylog<log::level::fatal>(logs area, const char *fmt, ...)
+void logger_init()
 {
-	va_list args;
-	va_start(args, fmt);
-	s_log.output(level::fatal, area, fmt, args);
-	va_end(args);
-	while(1) __asm__ ("hlt");
+	new (&g_logger) log::logger(log_buffer, sizeof(log_buffer));
+	g_logger.set_immediate(output_log);
 }
-
-template <log::level L>
-void
-log::trylog(logs area, const char *fmt, ...)
-{
-	auto i = static_cast<std::underlying_type<level>::type>(L);
-	if ((s_log.m_enabled[i] & bit_mask(area)) == 0) return;
-
-	va_list args;
-	va_start(args, fmt);
-	s_log.output(L, area, fmt, args);
-	va_end(args);
-}
-
-void
-log::output(level severity, logs area, const char *fmt, va_list args)
-{
-	m_cons->set_color(level_colors[static_cast<int>(severity)]);
-
-	m_cons->printf("%s %s: ",
-			areas[static_cast<int>(area)],
-			levels[static_cast<int>(severity)]);
-	m_cons->vprintf(fmt, args);
-	m_cons->set_color();
-	m_cons->puts("\n");
-}
-
-const log::trylog_p log::debug = &trylog<level::debug>;
-const log::trylog_p log::info = &trylog<level::info>;
-const log::trylog_p log::warn = &trylog<level::warn>;
-const log::trylog_p log::error = &trylog<level::error>;
-const log::trylog_p log::fatal = &trylog<level::fatal>;

src/kernel/log.h

@@ -1,49 +1,9 @@
 #pragma once
-#include <stdarg.h>
-#include <stdint.h>
 
-class console;
+#include "kutil/logger.h"
 
+namespace log = kutil::log;
+namespace logs = kutil::logs;
 
-enum class logs
-{
-	boot,
-	memory,
-	apic,
-	device,
-	driver,
-
-	max
-};
-
-
-class log
-{
-public:
-	enum class level {debug, info, warn, error, fatal, max};
-
-	static void init(console *cons);
-	static void enable(logs type, level at_level);
-
-	template <level L>
-	static void trylog(logs area, const char *fmt, ...);
-	using trylog_p = void (*)(logs area, const char *fmt, ...);
-
-	static const trylog_p debug;
-	static const trylog_p info;
-	static const trylog_p warn;
-	static const trylog_p error;
-	static const trylog_p fatal;
-
-private:
-	void output(level severity, logs area, const char *fmt, va_list args);
-
-	/// Bitmasks for what categories are enabled. fatal is
-	/// always enabled, so leave it out.
-	uint64_t m_enabled[static_cast<uint64_t>(level::max) - 1];
-	console *m_cons;
-
-	log();
-	log(console *cons);
-	static log s_log;
-};
+void logger_init();
+void logger_task();

src/kernel/main.cpp

@@ -1,124 +1,142 @@
 #include <stddef.h>
 #include <stdint.h>
 
+#include "initrd/initrd.h"
 #include "kutil/assert.h"
-#include "kutil/memory.h"
-
-#include "ahci/driver.h"
-#include "ahci/port.h"
-
+#include "apic.h"
+#include "block_device.h"
 #include "console.h"
 #include "cpu.h"
 #include "device_manager.h"
-#include "font.h"
+#include "gdt.h"
 #include "interrupts.h"
 #include "io.h"
-#include "kernel_data.h"
+#include "kernel_args.h"
 #include "log.h"
-#include "memory.h"
 #include "page_manager.h"
-#include "screen.h"
+#include "scheduler.h"
 #include "serial.h"
+#include "syscall.h"
 
 extern "C" {
-	void do_the_set_registers(popcorn_data *header);
-	void kernel_main(popcorn_data *header);
-
+	void kernel_main(kernel_args *header);
 	void *__bss_start, *__bss_end;
 }
 
-extern ahci_driver ahcid;
-
 extern void __kernel_assert(const char *, unsigned, const char *);
 
 void
-init_console(const popcorn_data *header)
+init_console()
 {
 	serial_port *com1 = new (&g_com1) serial_port(COM1);
 	console *cons = new (&g_console) console(com1);
 
-	if (header->frame_buffer) {
-		screen *s = new screen(
-				header->frame_buffer,
-				header->hres,
-				header->vres,
-				header->rmask,
-				header->gmask,
-				header->bmask);
-		font *f = new font(header->font);
-		cons->init_screen(s, f);
-	}
-
 	cons->set_color(0x21, 0x00);
 	cons->puts("Popcorn OS ");
 	cons->set_color(0x08, 0x00);
 	cons->puts(GIT_VERSION " booting...\n");
 
-	log::init(cons);
-	log::enable(logs::apic, log::level::info);
-	log::enable(logs::device, log::level::debug);
-	log::enable(logs::driver, log::level::debug);
-	log::enable(logs::memory, log::level::debug);
+	logger_init();
 }
 
-void do_error_3() { volatile int x = 1; volatile int y = 0; volatile int z = x / y; }
-void do_error_2() { do_error_3(); }
-void do_error_1() { do_error_2(); }
-
 void
-kernel_main(popcorn_data *header)
+kernel_main(kernel_args *header)
 {
+	bool waiting = header && (header->flags && POPCORN_FLAG_DEBUG);
+	while (waiting);
+
 	kutil::assert_set_callback(__kernel_assert);
 
-	page_manager *pager = new (&g_page_manager) page_manager;
+	gdt_init();
+	interrupts_init();
 
-	memory_initialize(
+	kutil::allocator &heap = memory_initialize(
+			header->scratch_pages,
 			header->memory_map,
 			header->memory_map_length,
 			header->memory_map_desc_size);
 
-	pager->map_offset_pointer(
-			&header->frame_buffer,
-			header->frame_buffer_length);
+	if (header->frame_buffer && header->frame_buffer_length) {
+		page_manager::get()->map_offset_pointer(
+				&header->frame_buffer,
+				header->frame_buffer_length);
+	}
 
-	init_console(header);
-	// pager->dump_blocks();
+	init_console();
 
-	interrupts_init();
+	log::debug(logs::boot, " Popcorn header is at: %016lx", header);
+	log::debug(logs::boot, "    Framebuffer is at: %016lx", header->frame_buffer);
+	log::debug(logs::boot, "    Kernel data is at: %016lx", header->data);
+	log::debug(logs::boot, "     Memory map is at: %016lx", header->memory_map);
+	log::debug(logs::boot, "ACPI root table is at: %016lx", header->acpi_table);
+	log::debug(logs::boot, "Runtime service is at: %016lx", header->runtime);
+
+	initrd::disk ird(header->initrd, heap);
+	log::info(logs::boot, "initrd loaded with %d files.", ird.files().count());
+	for (auto &f : ird.files())
+		log::info(logs::boot, "  %s%s (%d bytes).", f.executable() ? "*" : "", f.name(), f.size());
+
+	/*
+	   page_manager::get()->dump_pml4(nullptr, 0);
+	   page_manager::get()->dump_blocks(true);
+	*/
 
 	device_manager *devices =
-		new (&device_manager::get()) device_manager(header->acpi_table);
+		new (&device_manager::get()) device_manager(header->acpi_table, heap);
 
 	interrupts_enable();
 
+	/*
 	cpu_id cpu;
 	log::info(logs::boot, "CPU Vendor: %s", cpu.vendor_id());
 	log::info(logs::boot, "CPU Family %x Model %x Stepping %x",
 			cpu.family(), cpu.model(), cpu.stepping());
 
+	auto r = cpu.get(0x15);
+	log::info(logs::boot, "CPU Crystal: %dHz", r.ecx);
+
+	uintptr_t cr4 = 0;
+	__asm__ __volatile__ ( "mov %%cr4, %0" : "=r" (cr4) );
+	log::info(logs::boot, "cr4: %016x", cr4);
+	*/
+
 	devices->init_drivers();
 
-	ahci::port *disk = ahcid.find_disk();
+	/*
+	block_device *disk = devices->get_block_device(0);
 	if (disk) {
-		uint8_t buf[512];
-		kutil::memset(buf, 0, 512);
+		for (int i=0; i<1; ++i) {
+			uint8_t buf[512];
+			kutil::memset(buf, 0, 512);
 
-		disk->read(1, sizeof(buf), buf);
-		while (buf[0] == 0) io_wait();
+			kassert(disk->read(0x200, sizeof(buf), buf),
+					"Disk read returned 0");
 
-		console *cons = console::get();
-		uint8_t *p = &buf[0];
-		for (int i = 0; i < 8; ++i) {
-			for (int j = 0; j < 16; ++j) {
-				cons->printf(" %02x", *p++);
+			console *cons = console::get();
+			uint8_t *p = &buf[0];
+			for (int i = 0; i < 8; ++i) {
+				for (int j = 0; j < 16; ++j) {
+					cons->printf(" %02x", *p++);
+				}
+				cons->putc('\n');
 			}
-			cons->putc('\n');
 		}
+	} else {
+		log::warn(logs::boot, "No block devices present.");
+	}
+	*/
+
+	devices->get_lapic()->calibrate_timer();
+
+	syscall_enable();
+	scheduler *sched = new (&scheduler::get()) scheduler(devices->get_lapic(), heap);
+
+	sched->create_kernel_task(-1, logger_task);
+
+	for (auto &f : ird.files()) {
+		if (f.executable())
+			sched->load_process(f.name(), f.data(), f.size());
 	}
 
-	// do_error_1();
-	// __asm__ __volatile__("int $15");
-
-	g_console.puts("boogity!");
-	do_the_set_registers(header);
+	sched->start();
 }

src/kernel/memory_bootstrap.cpp

@@ -1,9 +1,27 @@
+#include <algorithm>
+#include <utility>
+#include "kutil/address_manager.h"
 #include "kutil/assert.h"
-#include "kutil/memory.h"
-#include "memory.h"
+#include "kutil/heap_allocator.h"
+#include "frame_allocator.h"
+#include "io.h"
+#include "log.h"
 #include "page_manager.h"
 
-const unsigned efi_page_size = 0x1000;
+using memory::frame_size;
+using memory::kernel_max_heap;
+using memory::kernel_offset;
+using memory::page_offset;
+
+static const unsigned ident_page_flags = 0xb;
+
+kutil::address_manager g_kernel_address_manager;
+kutil::heap_allocator g_kernel_heap;
+
+void * operator new(size_t size)           { return g_kernel_heap.allocate(size); }
+void * operator new [] (size_t size)       { return g_kernel_heap.allocate(size); }
+void operator delete (void *p) noexcept    { return g_kernel_heap.free(p); }
+void operator delete [] (void *p) noexcept { return g_kernel_heap.free(p); }
 
 enum class efi_memory_type : uint32_t
 {
@@ -26,66 +44,13 @@ enum class efi_memory_type : uint32_t
 	efi_max,
 
 	popcorn_kernel = 0x80000000,
-	popcorn_font,
 	popcorn_data,
-	popcorn_log,
-	popcorn_pml4,
+	popcorn_initrd,
+	popcorn_scratch,
 
 	popcorn_max
 };
 
-const char *efi_memory_type_names[] = {
-	"             reserved",
-	"          loader_code",
-	"          loader_data",
-	"   boot_services_code",
-	"   boot_services_data",
-	"runtime_services_code",
-	"runtime_services_data",
-	"            available",
-	"             unusable",
-	"         acpi_reclaim",
-	"             acpi_nvs",
-	"                 mmio",
-	"            mmio_port",
-	"             pal_code",
-
-	"       popcorn_kernel",
-	"         popcorn_font",
-	"         popcorn_data",
-	"          popcorn_log",
-	"         popcorn_pml4",
-};
-
-static const char *
-get_efi_name(efi_memory_type t)
-{
-	static const unsigned offset =
-		(unsigned)efi_memory_type::popcorn_kernel - (unsigned)efi_memory_type::efi_max;
-
-	return t >= efi_memory_type::popcorn_kernel ?
-		efi_memory_type_names[(unsigned)t - offset] :
-		efi_memory_type_names[(unsigned)t];
-}
-
-enum class efi_memory_flag : uint64_t
-{
-	can_mark_uc  = 0x0000000000000001, // uc = un-cacheable
-	can_mark_wc  = 0x0000000000000002, // wc = write-combining
-	can_mark_wt  = 0x0000000000000004, // wt = write through
-	can_mark_wb  = 0x0000000000000008, // wb = write back
-	can_mark_uce = 0x0000000000000010, // uce = un-cacheable exported
-	can_mark_wp  = 0x0000000000001000, // wp = write protected
-	can_mark_rp  = 0x0000000000002000, // rp = read protected
-	can_mark_xp  = 0x0000000000004000, // xp = exceute protected
-	can_mark_ro  = 0x0000000000020000, // ro = read only
-
-	non_volatile  = 0x0000000000008000,
-	more_reliable = 0x0000000000010000,
-	runtime       = 0x8000000000000000
-};
-IS_BITFIELD(efi_memory_flag);
-
 struct efi_memory_descriptor
 {
 	efi_memory_type type;
@@ -93,440 +58,185 @@ struct efi_memory_descriptor
 	uint64_t physical_start;
 	uint64_t virtual_start;
 	uint64_t pages;
-	efi_memory_flag flags;
+	uint64_t flags;
 };
 
-static const efi_memory_descriptor *
-desc_incr(const efi_memory_descriptor *d, size_t desc_length)
+struct memory_map
 {
-	return reinterpret_cast<const efi_memory_descriptor *>(
-			reinterpret_cast<const uint8_t *>(d) + desc_length);
-}
+	memory_map(const void *efi_map, size_t map_length, size_t desc_length) :
+		efi_map(efi_map), map_length(map_length), desc_length(desc_length) {}
 
-static unsigned
-count_table_pages_needed(page_block *used)
-{
-	page_table_indices last_idx{~0ull};
-	unsigned counts[] = {1, 0, 0, 0};
+	class iterator
+	{
+	public:
+		iterator(const memory_map &map, efi_memory_descriptor const *item) :
+			map(map), item(item) {}
 
-	for (page_block *cur = used; cur; cur = cur->next) {
-		if (!cur->has_flag(page_block_flags::mapped))
-			continue;
-
-		page_table_indices start{cur->virtual_address};
-		page_table_indices end{cur->virtual_address + (cur->count * page_manager::page_size)};
-
-		counts[1] +=
-			((start[0] == last_idx[0]) ? 0 : 1) +
-			(end[0] - start[0]);
-
-		counts[2] +=
-			((start[0] == last_idx[0] &&
-			 start[1] == last_idx[1]) ? 0 : 1) +
-			(end[1] - start[1]);
-
-		counts[3] +=
-			((start[0] == last_idx[0] &&
-			 start[1] == last_idx[1] &&
-			 start[2] == last_idx[2]) ? 0 : 1) +
-			(end[2] - start[2]);
-
-		last_idx = end;
-	}
-
-	return counts[0] + counts[1] + counts[2] + counts[3];
-
-}
-
-page_block *
-remove_block_for(page_block **list, uint64_t phys_start, uint64_t pages, page_block **cache)
-{
-	// This is basically just the removal portion of page_manager::unmap_pages,
-	// but with physical addresses, and only ever removing a single block.
-
-	page_block *prev = nullptr;
-	page_block *cur = *list;
-	while (cur && !cur->contains_physical(phys_start)) {
-		prev = cur;
-		cur = cur->next;
-	}
-
-	kassert(cur, "Couldn't find block to remove");
-
-	uint64_t size = page_manager::page_size * pages;
-	uint64_t end = phys_start + size;
-	uint64_t leading = phys_start - cur->physical_address;
-	uint64_t trailing = cur->physical_end() - end;
-
-	if (leading) {
-		uint64_t pages = leading / page_manager::page_size;
-
-		page_block *lead_block = *cache;
-		*cache = (*cache)->next;
-
-		lead_block->copy(cur);
-		lead_block->next = cur;
-		lead_block->count = pages;
-
-		cur->count -= pages;
-		cur->physical_address += leading;
-
-		if (cur->virtual_address)
-			cur->virtual_address += leading;
-
-		if (prev) {
-			prev->next = lead_block;
-		} else {
-			prev = lead_block;
-			*list = prev;
-		}
-	}
-
-	if (trailing) {
-		uint64_t pages = trailing / page_manager::page_size;
-
-		page_block *trail_block = *cache;
-		*cache = (*cache)->next;
-
-		trail_block->copy(cur);
-		trail_block->next = cur->next;
-		trail_block->count = pages;
-		trail_block->physical_address += size;
-
-		if (cur->virtual_address)
-			trail_block->virtual_address += size;
-
-		cur->count -= pages;
-		cur->next = trail_block;
-	}
-
-	prev->next = cur->next;
-	cur->next = nullptr;
-	return cur;
-}
-
-uint64_t
-gather_block_lists(
-		uint64_t scratch_virt,
-		const void *memory_map,
-		size_t map_length,
-		size_t desc_length,
-		page_block **free_head,
-		page_block **used_head)
-{
-	int i = 0;
-	page_block *free = nullptr;
-	page_block *used = nullptr;
-
-	page_block *block_list = reinterpret_cast<page_block *>(scratch_virt);
-	efi_memory_descriptor const *desc = reinterpret_cast<efi_memory_descriptor const *>(memory_map);
-	efi_memory_descriptor const *end = desc_incr(desc, map_length);
-
-	while (desc < end) {
-		page_block *block = &block_list[i++];
-		block->physical_address = desc->physical_start;
-		block->virtual_address = desc->virtual_start;
-		block->count = desc->pages;
-		block->next = nullptr;
-
-		switch (desc->type) {
-		case efi_memory_type::loader_code:
-		case efi_memory_type::loader_data:
-			block->flags = page_block_flags::used | page_block_flags::pending_free;
-			break;
-
-		case efi_memory_type::boot_services_code:
-		case efi_memory_type::boot_services_data:
-		case efi_memory_type::available:
-			block->flags = page_block_flags::free;
-			break;
-
-		case efi_memory_type::acpi_reclaim:
-			block->flags =
-				page_block_flags::used |
-				page_block_flags::mapped |
-				page_block_flags::acpi_wait;
-
-			block->virtual_address = block->physical_address;
-			break;
-
-		case efi_memory_type::persistent:
-			block->flags = page_block_flags::nonvolatile;
-			break;
-
-		default:
-			block->flags = page_block_flags::used | page_block_flags::permanent;
-			break;
+		inline efi_memory_descriptor const * operator*() const { return item; }
+		inline bool operator!=(const iterator &other) { return item != other.item; }
+		inline iterator & operator++() {
+			item = kutil::offset_pointer(item, map.desc_length);
+			return *this;
 		}
 
-		if (block->has_flag(page_block_flags::used)) {
-			if (block->virtual_address || !block->physical_address)
-				block->flags |= page_block_flags::mapped;
+	private:
+		const memory_map &map;
+		efi_memory_descriptor const *item;
+	};
 
-			used = page_block::insert(used, block);
-		} else {
-			free = page_block::insert(free, block);
-		}
-
-		desc = desc_incr(desc, desc_length);
+	iterator begin() const {
+		return iterator(*this, reinterpret_cast<efi_memory_descriptor const *>(efi_map));
 	}
 
-	*free_head = free;
-	*used_head = used;
-	return reinterpret_cast<uint64_t>(&block_list[i]);
-}
-
-page_block *
-fill_page_with_blocks(uint64_t start)
-{
-	uint64_t end = page_align(start);
-	uint64_t count = (end - start) / sizeof(page_block);
-	if (count == 0) return nullptr;
-
-	page_block *blocks = reinterpret_cast<page_block *>(start);
-	for (unsigned i = 0; i < count; ++i)
-		blocks[i].zero(&blocks[i+1]);
-	blocks[count - 1].next = nullptr;
-	return blocks;
-}
-
-void
-copy_new_table(page_table *base, unsigned index, page_table *new_table)
-{
-	uint64_t entry = base->entries[index];
-
-	// If this is a large page and not a a table, bail out.
-	if(entry & 0x80) return;
-
-	if (entry & 0x1) {
-		page_table *old_next = reinterpret_cast<page_table *>(
-				base->entries[index] & ~0xffful);
-		for (int i = 0; i < 512; ++i) new_table->entries[i] = old_next->entries[i];
-	} else {
-		for (int i = 0; i < 512; ++i) new_table->entries[i] = 0;
+	iterator end() const {
+		const void *end = kutil::offset_pointer(efi_map, map_length);
+		return iterator(*this, reinterpret_cast<efi_memory_descriptor const *>(end));
 	}
 
-	base->entries[index] = reinterpret_cast<uint64_t>(new_table) | 0xb;
-}
+	const void *efi_map;
+	size_t map_length;
+	size_t desc_length;
+};
 
-static uint64_t
-find_efi_free_aligned_pages(const void *memory_map, size_t map_length, size_t desc_length, unsigned pages)
+class memory_bootstrap
 {
-	efi_memory_descriptor const *desc =
-		reinterpret_cast<efi_memory_descriptor const *>(memory_map);
-	efi_memory_descriptor const *end = desc_incr(desc, map_length);
+public:
+	memory_bootstrap(const void *memory_map, size_t map_length, size_t desc_length) :
+		map(memory_map, map_length, desc_length) {}
 
-	const unsigned want_space = pages * page_manager::page_size;
-	uint64_t start_phys = 0;
-
-	for (; desc < end; desc = desc_incr(desc, desc_length)) {
-		if (desc->type != efi_memory_type::available)
-			continue;
-
-		// See if the first wanted pages fit in one page table. If we
-		// find free memory at zero, skip ahead because we're not ready
-		// to deal with 0 being a valid pointer yet.
-		start_phys = desc->physical_start;
-		if (start_phys == 0)
-			start_phys += efi_page_size;
-
-		const uint64_t desc_end =
-			desc->physical_start + desc->pages * efi_page_size;
-
-		uint64_t end = start_phys + want_space;
-		if (end < desc_end) {
-			page_table_indices start_idx{start_phys};
-			page_table_indices end_idx{end};
-			if (start_idx[0] == end_idx[0] &&
-				start_idx[1] == end_idx[1] &&
-				start_idx[2] == end_idx[2])
-				break;
-
-			// Try seeing if the page-table-aligned version fits
-			start_phys = page_table_align(start_phys);
-			end = start_phys + want_space;
-			if (end < desc_end)
-				break;
-		}
-	}
-
-	kassert(desc < end, "Couldn't find wanted pages of aligned scratch space.");
-	return start_phys;
-}
-
-static unsigned
-check_needs_page_ident(page_table *table, unsigned index, page_table **free_pages)
-{
-	if ((table->entries[index] & 0x1) == 1) return 0;
-
-	kassert(*free_pages, "check_needs_page_ident needed to allocate but had no free pages");
-
-	page_table *new_table = (*free_pages)++;
-	for (int i=0; i<512; ++i) new_table->entries[i] = 0;
-	table->entries[index] = reinterpret_cast<uint64_t>(new_table) | 0xb;
-	return 1;
-}
-
-static unsigned
-page_in_ident(
-		page_table *pml4,
-		uint64_t phys_addr,
-		uint64_t virt_addr,
-		uint64_t count,
-		page_table *free_pages)
-{
-	page_table_indices idx{virt_addr};
-	page_table *tables[4] = {pml4, nullptr, nullptr, nullptr};
-
-	unsigned pages_consumed = 0;
-	for (; idx[0] < 512; idx[0] += 1) {
-		pages_consumed += check_needs_page_ident(tables[0], idx[0], &free_pages);
-		tables[1] = reinterpret_cast<page_table *>(
-				tables[0]->entries[idx[0]] & ~0xfffull);
-
-		for (; idx[1] < 512; idx[1] += 1, idx[2] = 0, idx[3] = 0) {
-			pages_consumed += check_needs_page_ident(tables[1], idx[1], &free_pages);
-			tables[2] = reinterpret_cast<page_table *>(
-					tables[1]->entries[idx[1]] & ~0xfffull);
-
-			for (; idx[2] < 512; idx[2] += 1, idx[3] = 0) {
-				pages_consumed += check_needs_page_ident(tables[2], idx[2], &free_pages);
-				tables[3] = reinterpret_cast<page_table *>(
-						tables[2]->entries[idx[2]] & ~0xfffull);
-
-				for (; idx[3] < 512; idx[3] += 1) {
-					tables[3]->entries[idx[3]] = phys_addr | 0xb;
-					phys_addr += page_manager::page_size;
-					if (--count == 0) return pages_consumed;
-				}
+	void add_free_frames(frame_allocator &fa) {
+		for (auto *desc : map) {
+			if (desc->type == efi_memory_type::loader_code ||
+				desc->type == efi_memory_type::loader_data ||
+				desc->type == efi_memory_type::boot_services_code ||
+				desc->type == efi_memory_type::boot_services_data ||
+				desc->type == efi_memory_type::available)
+			{
+				fa.free(desc->physical_start, desc->pages);
 			}
 		}
 	}
 
-	kassert(0, "Ran to end of page_in_ident");
-	return 0; // Cannot reach
-}
-
-void
-memory_initialize(const void *memory_map, size_t map_length, size_t desc_length)
-{
-	// The bootloader reserved 16 pages for page tables, which we'll use to bootstrap.
-	// The first one is the already-installed PML4, so grab it from CR3.
-	uint64_t cr3;
-	__asm__ __volatile__ ( "mov %%cr3, %0" : "=r" (cr3) );
-	page_table *tables = reinterpret_cast<page_table *>(cr3 & ~0xfffull);
-
-	// Now go through EFi's memory map and find a region of scratch space.
-	const unsigned want_pages = 32;
-	uint64_t free_region_start_phys =
-		find_efi_free_aligned_pages(memory_map, map_length, desc_length, want_pages);
-
-	// Offset-map this region into the higher half.
-	uint64_t free_region_start_virt =
-		free_region_start_phys + page_manager::high_offset;
-
-	uint64_t free_next = free_region_start_virt;
-
-	// We'll need to copy any existing tables (except the PML4 which the
-	// bootloader gave us) into our 4 reserved pages so we can edit them.
-	page_table_indices fr_idx{free_region_start_virt};
-
-	copy_new_table(&tables[0], fr_idx[0], &tables[1]);
-	copy_new_table(&tables[1], fr_idx[1], &tables[2]);
-	copy_new_table(&tables[2], fr_idx[2], &tables[3]);
-	page_in_ident(&tables[0], free_region_start_phys, free_region_start_virt, want_pages, nullptr);
-
-	// We now have pages starting at "free_next" to bootstrap ourselves. Start by
-	// taking inventory of free pages.
-	page_block *free_head = nullptr;
-	page_block *used_head = nullptr;
-	free_next = gather_block_lists(
-			free_next, memory_map, map_length, desc_length,
-			&free_head, &used_head);
-
-	// Unused page_block structs go here - finish out the current page with them
-	page_block *cache_head = fill_page_with_blocks(free_next);
-	free_next = page_align(free_next);
-
-	// Now go back through these lists and consolidate
-	page_block *freed = page_block::consolidate(free_head);
-	cache_head = page_block::append(cache_head, freed);
-
-	freed = page_block::consolidate(used_head);
-	cache_head = page_block::append(cache_head, freed);
-
-
-	// Pull out the block that represents the bootstrap pages we've used
-	uint64_t used = free_next - free_region_start_virt;
-	uint64_t used_pages = used / page_manager::page_size;
-	uint64_t remaining_pages = want_pages - used_pages;
-
-	page_block *removed = remove_block_for(
-			&free_head,
-			free_region_start_phys,
-			used_pages,
-			&cache_head);
-
-	kassert(removed, "remove_block_for didn't find the bootstrap region.");
-	kassert(removed->physical_address == free_region_start_phys,
-			"remove_block_for found the wrong region.");
-
-	// Add it to the used list
-	removed->virtual_address = free_region_start_virt;
-	removed->flags = page_block_flags::used | page_block_flags::mapped;
-	used_head = page_block::insert(used_head, removed);
-
-	// Pull out the block that represents the rest
-	uint64_t free_next_phys = free_region_start_phys + used;
-
-	removed = remove_block_for(
-			&free_head,
-			free_next_phys,
-			remaining_pages,
-			&cache_head);
-
-	kassert(removed, "remove_block_for didn't find the page table region.");
-	kassert(removed->physical_address == free_next_phys,
-			"remove_block_for found the wrong region.");
-
-	uint64_t pt_start_phys = removed->physical_address;
-	uint64_t pt_start_virt = removed->physical_address + page_manager::page_offset;
-
-	// Record that we're about to remap it into the page table address space
-	removed->virtual_address = pt_start_virt;
-	removed->flags = page_block_flags::used | page_block_flags::mapped;
-	used_head = page_block::insert(used_head, removed);
-
-	page_manager *pm = &g_page_manager;
-
-	// Actually remap them into page table space
-	pm->page_out(&tables[0], free_next, remaining_pages);
-
-	page_table_indices pg_idx{pt_start_virt};
-	copy_new_table(&tables[0], pg_idx[0], &tables[4]);
-	copy_new_table(&tables[4], pg_idx[1], &tables[5]);
-	copy_new_table(&tables[5], pg_idx[2], &tables[6]);
-
-	page_in_ident(&tables[0], pt_start_phys, pt_start_virt, remaining_pages, tables + 4);
-
-	// Finally, build an acutal set of kernel page tables that just contains
-	// what the kernel actually has mapped, but making everything writable
-	// (especially the page tables themselves)
-	page_table *pml4 = reinterpret_cast<page_table *>(pt_start_virt);
-	for (int i=0; i<512; ++i) pml4->entries[i] = 0;
-
-	// Give the rest to the page_manager's cache for use in page_in
-	pm->free_table_pages(pml4 + 1, remaining_pages - 1);
-
-	for (page_block *cur = used_head; cur; cur = cur->next) {
-		if (!cur->has_flag(page_block_flags::mapped)) continue;
-		pm->page_in(pml4, cur->physical_address, cur->virtual_address, cur->count);
+	void add_used_frames(kutil::address_manager &am) {
+		for (auto *desc : map) {
+			if (desc->type == efi_memory_type::popcorn_data ||
+				desc->type == efi_memory_type::popcorn_initrd)
+			{
+				uintptr_t virt_addr = desc->physical_start + kernel_offset;
+				am.mark(virt_addr, desc->pages * frame_size);
+			}
+			else if (desc->type == efi_memory_type::popcorn_kernel)
+			{
+				uintptr_t virt_addr = desc->physical_start + kernel_offset;
+				am.mark_permanent(virt_addr, desc->pages * frame_size);
+			}
+		}
 	}
 
-	// Put our new PML4 into CR3 to start using it
-	page_manager::set_pml4(pml4);
+	void page_in_kernel(page_manager &pm, page_table *pml4) {
+		for (auto *desc : map) {
+			if (desc->type == efi_memory_type::popcorn_kernel ||
+				desc->type == efi_memory_type::popcorn_data ||
+				desc->type == efi_memory_type::popcorn_initrd)
+			{
+				uintptr_t virt_addr = desc->physical_start + kernel_offset;
+				pm.page_in(pml4, desc->physical_start, virt_addr, desc->pages);
+			}
 
-	// We now have all used memory mapped ourselves. Let the page_manager take
-	// over from here.
-	g_page_manager.init(free_head, used_head, cache_head);
+			if (desc->type == efi_memory_type::acpi_reclaim) {
+				pm.page_in(pml4, desc->physical_start, desc->physical_start, desc->pages);
+			}
+		}
+
+		// Put our new PML4 into CR3 to start using it
+		page_manager::set_pml4(pml4);
+		pm.m_kernel_pml4 = pml4;
+	}
+
+private:
+	const memory_map map;
+};
+
+kutil::allocator &
+memory_initialize(uint16_t scratch_pages, const void *memory_map, size_t map_length, size_t desc_length)
+{
+	// make sure the options we want in CR4 are set
+	uint64_t cr4;
+	__asm__ __volatile__ ( "mov %%cr4, %0" : "=r" (cr4) );
+	cr4 |=
+		0x000080 | // Enable global pages
+		0x000200 | // Enable FXSAVE/FXRSTOR
+		0x010000 | // Enable FSGSBASE
+		0x020000 | // Enable PCIDs
+		0;
+	__asm__ __volatile__ ( "mov %0, %%cr4" :: "r" (cr4) );
+
+	// The bootloader reserved "scratch_pages" pages for page tables and
+	// scratch space, which we'll use to bootstrap.  The first one is the
+	// already-installed PML4, so grab it from CR3.
+	uint64_t scratch_phys;
+	__asm__ __volatile__ ( "mov %%cr3, %0" : "=r" (scratch_phys) );
+	scratch_phys &= ~0xfffull;
+
+	// The tables are ident-mapped currently, so the cr3 physical address works. But let's
+	// get them into the offset-mapped area asap.
+	page_table *tables = reinterpret_cast<page_table *>(scratch_phys);
+
+	page_table *id_pml4 = &tables[0];
+	page_table *id_pdp = &tables[1];
+	for (int i=0; i<512; ++i)
+		id_pdp->entries[i] = (static_cast<uintptr_t>(i) << 30) | 0x18b;
+	id_pml4->entries[511] = reinterpret_cast<uintptr_t>(id_pdp) | 0x10b;
+
+	// Make sure the page table is finished updating before we write to memory
+	__sync_synchronize();
+	io_wait();
+
+	uintptr_t scratch_virt = scratch_phys + page_offset;
+	memory_bootstrap bootstrap {memory_map, map_length, desc_length};
+
+	// Now tell the frame allocator what's free
+	frame_allocator *fa = new (&g_frame_allocator) frame_allocator;
+	bootstrap.add_free_frames(*fa);
+
+	// Build an initial address manager that we'll copy into the real
+	// address manager later (so that we can use a raw allocator now)
+	kutil::allocator &alloc = fa->raw_allocator();
+	kutil::address_manager init_am(alloc);
+	init_am.add_regions(kernel_offset, page_offset - kernel_offset);
+	bootstrap.add_used_frames(init_am);
+
+	// Add the heap into the address manager
+	uintptr_t heap_start = page_offset - kernel_max_heap;
+	init_am.mark(heap_start, kernel_max_heap);
+
+	kutil::allocator *heap_alloc =
+		new (&g_kernel_heap) kutil::heap_allocator(heap_start, kernel_max_heap);
+
+	// Copy everything into the real address manager
+	kutil::address_manager *am =
+		new (&g_kernel_address_manager) kutil::address_manager(
+				std::move(init_am), *heap_alloc);
+
+	// Create the page manager
+	page_manager *pm = new (&g_page_manager) page_manager(*fa, *am);
+
+	// Give the frame_allocator back the rest of the scratch pages
+	fa->free(scratch_phys + (3 * frame_size), scratch_pages - 3);
+
+	// Finally, build an acutal set of kernel page tables where we'll only add
+	// what the kernel actually has mapped, but making everything writable
+	// (especially the page tables themselves)
+	page_table *pml4 = &tables[2];
+	pml4 = kutil::offset_pointer(pml4, page_offset);
+
+	kutil::memset(pml4, 0, sizeof(page_table));
+	pml4->entries[511] = reinterpret_cast<uintptr_t>(id_pdp) | 0x10b;
+
+	bootstrap.page_in_kernel(*pm, pml4);
+
+	// Reclaim the old PML4
+	fa->free(scratch_phys, 1);
+
+	return *heap_alloc;
 }

src/kernel/msr.cpp

@@ -0,0 +1,18 @@
+#include "msr.h"
+
+uint64_t
+rdmsr(msr addr)
+{
+	uint32_t low, high;
+	__asm__ __volatile__ ("rdmsr" : "=a"(low), "=d"(high) : "c"(addr));
+	return (static_cast<uint64_t>(high) << 32) | low;
+}
+
+void
+wrmsr(msr addr, uint64_t value)
+{
+	uint32_t low = value & 0xffffffff;
+	uint32_t high = value >> 32;
+	__asm__ __volatile__ ("wrmsr" :: "c"(addr), "a"(low), "d"(high));
+}
+

src/kernel/msr.h

@@ -0,0 +1,27 @@
+#pragma once
+/// \file msr.h
+/// Routines and definitions for dealing with Model-Specific Registers
+
+#include <stdint.h>
+
+enum class msr : uint32_t
+{
+	ia32_efer			= 0xc0000080,
+	ia32_star			= 0xc0000081,
+	ia32_lstar			= 0xc0000082,
+	ia32_fmask			= 0xc0000084,
+
+	ia32_gs_base		= 0xc0000101,
+	ia32_kernel_gs_base = 0xc0000102
+};
+
+/// Read the value of a MSR
+/// \arg addr   The MSR address
+/// \returns    The current value of the MSR
+uint64_t rdmsr(msr addr);
+
+/// Write to a MSR
+/// \arg addr   The MSR address
+/// \arg value  The value to write
+void wrmsr(msr addr, uint64_t value);
+

src/kernel/page_manager.cpp

@@ -1,24 +1,33 @@
 #include <algorithm>
 
 #include "kutil/assert.h"
-#include "kutil/memory_manager.h"
+#include "console.h"
+#include "io.h"
 #include "log.h"
 #include "page_manager.h"
 
-page_manager g_page_manager;
+using memory::frame_size;
+using memory::kernel_offset;
+using memory::page_offset;
+using memory::page_mappable;
+
+extern kutil::address_manager g_kernel_address_manager;
+page_manager g_page_manager(
+	g_frame_allocator,
+	g_kernel_address_manager);
 
 
-static addr_t
+static uintptr_t
 pt_to_phys(page_table *pt)
 {
-	return reinterpret_cast<addr_t>(pt) - page_manager::page_offset;
+	return reinterpret_cast<uintptr_t>(pt) - page_offset;
 }
 
 
 static page_table *
-pt_from_phys(addr_t p)
+pt_from_phys(uintptr_t p)
 {
-	return reinterpret_cast<page_table *>((p + page_manager::page_offset) & ~0xfffull);
+	return reinterpret_cast<page_table *>((p + page_offset) & ~0xfffull);
 }
 
 
@@ -29,300 +38,204 @@ struct free_page_header
 };
 
 
-void mm_grow_callback(void *next, size_t length)
+page_manager::page_manager(
+		frame_allocator &frames,
+		kutil::address_manager &addrs) :
+	m_page_cache(nullptr),
+	m_frames(frames),
+	m_addrs(addrs)
 {
-	kassert(length % page_manager::page_size == 0,
-			"Heap manager requested a fractional page.");
-
-	size_t pages = length / page_manager::page_size;
-	log::info(logs::memory, "Heap manager growing heap by %d pages.", pages);
-	g_page_manager.map_pages(reinterpret_cast<addr_t>(next), pages);
 }
 
-
-size_t
-page_block::length(page_block *list)
+page_table *
+page_manager::create_process_map()
 {
-	size_t i = 0;
-	for (page_block *b = list; b; b = b->next) ++i;
-	return i;
+	page_table *table = get_table_page();
+
+	kutil::memset(table, 0, frame_size);
+	table->entries[510] = m_kernel_pml4->entries[510];
+	table->entries[511] = m_kernel_pml4->entries[511];
+
+	// Create the initial user stack
+	map_pages(
+		memory::initial_stack - (memory::initial_stack_pages * frame_size),
+		memory::initial_stack_pages,
+		true, // This is the ring3 stack, user = true
+		table);
+
+	return table;
 }
 
-page_block *
-page_block::append(page_block *list, page_block *extra)
+uintptr_t
+page_manager::copy_page(uintptr_t orig)
 {
-	if (list == nullptr) return extra;
-	else if (extra == nullptr) return list;
+	bool paged_orig = false;
+	bool paged_copy = false;
 
-	page_block *cur = list;
-	while (cur->next)
-		cur = cur->next;
-	cur->next = extra;
-	return list;
-}
+	uintptr_t orig_virt;
 
-page_block *
-page_block::insert(page_block *list, page_block *block)
-{
-	if (list == nullptr) return block;
-	else if (block == nullptr) return list;
-
-	page_block *cur = list;
-	page_block *prev = nullptr;
-	while (cur && page_block::compare(block, cur) > 0) {
-		prev = cur;
-		cur = cur->next;
+	if (page_mappable(orig)) {
+		orig_virt = orig + page_offset;
+	} else {
+		orig_virt = m_addrs.allocate(frame_size);
+		page_in(get_pml4(), orig, orig_virt, 1);
+		paged_orig = true;
 	}
 
-	block->next = cur;
-	if (prev) {
-		prev->next = block;
-		return list;
+	uintptr_t copy = 0;
+	uintptr_t copy_virt;
+	size_t n = m_frames.allocate(1, &copy);
+	kassert(n, "copy_page could not allocate page");
+
+	if (page_mappable(copy)) {
+		copy_virt = copy + page_offset;
+	} else {
+		copy_virt = m_addrs.allocate(frame_size);
+		page_in(get_pml4(), copy, copy_virt, 1);
+		paged_copy = true;
 	}
-	return block;
+
+	// TODO: multiple page copies at a time, so that we don't have to keep
+	// paying this mapping penalty
+	if (paged_orig || paged_copy) {
+		set_pml4(get_pml4());
+		__sync_synchronize();
+		io_wait();
+	}
+
+	kutil::memcpy(
+			reinterpret_cast<void *>(copy_virt),
+			reinterpret_cast<void *>(orig_virt),
+			frame_size);
+
+	if (paged_orig) {
+		page_out(get_pml4(), orig_virt, 1);
+		m_addrs.free(orig_virt);
+	}
+
+	if (paged_copy) {
+		page_out(get_pml4(), copy_virt, 1);
+		m_addrs.free(copy_virt);
+	}
+
+	return copy;
 }
 
-int
-page_block::compare(const page_block *lhs, const page_block *rhs)
+page_table *
+page_manager::copy_table(page_table *from, page_table::level lvl, page_table_indices index)
 {
-	if (lhs->virtual_address < rhs->virtual_address)
-		return -1;
-	else if (lhs->virtual_address > rhs->virtual_address)
-		return 1;
+	page_table *to = get_table_page();
+	log::debug(logs::paging, "Page manager copying level %d table at %016lx to %016lx.", lvl, from, to);
 
-	if (lhs->physical_address < rhs->physical_address)
-		return -1;
-	else if (lhs->physical_address > rhs->physical_address)
-		return 1;
+	if (lvl == page_table::level::pml4) {
+		to->entries[510] = m_kernel_pml4->entries[510];
+		to->entries[511] = m_kernel_pml4->entries[511];
+	}
 
-	return 0;
-}
+	const int max =
+		lvl == page_table::level::pml4 ?
+			510 :
+			512;
 
-page_block *
-page_block::consolidate(page_block *list)
-{
-	page_block *freed = nullptr;
-	page_block *cur = list;
+	unsigned pages_copied = 0;
+	uintptr_t from_addr = 0;
+	uintptr_t to_addr = 0;
 
-	while (cur) {
-		page_block *next = cur->next;
-
-		if (next &&
-			cur->flags == next->flags &&
-			cur->physical_end() == next->physical_address &&
-			(!cur->has_flag(page_block_flags::mapped) ||
-			cur->virtual_end() == next->virtual_address)) {
-
-			cur->count += next->count;
-			cur->next = next->next;
-
-			next->zero(freed);
-			freed = next;
+	for (int i = 0; i < max; ++i) {
+		if (!from->is_present(i)) {
+			to->entries[i] = 0;
 			continue;
 		}
 
-		cur = cur->next;
-	}
+		index[lvl] = i;
 
-	return freed;
-}
+		bool is_page =
+			lvl == page_table::level::pt ||
+			from->is_large_page(lvl, i);
 
-void
-page_block::dump(page_block *list, const char *name, bool show_unmapped)
-{
-	log::info(logs::memory, "Block list %s:", name);
-
-	int count = 0;
-	for (page_block *cur = list; cur; cur = cur->next) {
-		count += 1;
-		if (!(show_unmapped || cur->has_flag(page_block_flags::mapped)))
-			continue;
-
-		if (cur->virtual_address) {
-			page_table_indices start{cur->virtual_address};
-			log::info(logs::memory, "  %lx %x [%6d] %lx (%d,%d,%d,%d)",
-					cur->physical_address,
-					cur->flags,
-					cur->count,
-					cur->virtual_address,
-					start[0], start[1], start[2], start[3]);
+		if (is_page) {
+			uint16_t flags = from->entries[i] &  0xfffull;
+			uintptr_t orig = from->entries[i] & ~0xfffull;
+			to->entries[i] = copy_page(orig) | flags;
+			if (!pages_copied++)
+				from_addr = index.addr();
+			to_addr = index.addr();
 		} else {
-			page_table_indices start{cur->virtual_address};
-			log::info(logs::memory, "  %lx %x [%6d]",
-					cur->physical_address,
-					cur->flags,
-					cur->count);
+			uint16_t flags = 0;
+			page_table *next_from = from->get(i, &flags);
+			page_table *next_to = copy_table(next_from, page_table::deeper(lvl), index);
+			to->set(i, next_to, flags);
 		}
 	}
 
-	log::info(logs::memory, "  Total: %d", count);
+	if (pages_copied)
+		log::debug(logs::paging, "   copied %3u pages %016lx - %016lx",
+			pages_copied, from_addr, to_addr + frame_size);
+
+	return to;
 }
 
 void
-page_block::zero(page_block *set_next)
+page_manager::delete_process_map(page_table *pml4)
 {
-	physical_address = 0;
-	virtual_address = 0;
-	count = 0;
-	flags = page_block_flags::free;
-	next = set_next;
-}
+	bool was_pml4 = (pml4 == get_pml4());
+	if (was_pml4)
+		set_pml4(m_kernel_pml4);
 
-void
-page_block::copy(page_block *other)
-{
-	physical_address = other->physical_address;
-	virtual_address = other->virtual_address;
-	count = other->count;
-	flags = other->flags;
-	next = other->next;
-}
+	log::info(logs::paging, "Deleting process pml4 at %016lx%s",
+			pml4, was_pml4 ? " (was current)" : "");
 
-
-page_manager::page_manager() :
-	m_free(nullptr),
-	m_used(nullptr),
-	m_block_cache(nullptr),
-	m_page_cache(nullptr)
-{
-	kassert(this == &g_page_manager, "Attempt to create another page_manager.");
-}
-
-void
-page_manager::init(
-	page_block *free,
-	page_block *used,
-	page_block *block_cache)
-{
-	m_free = free;
-	m_used = used;
-	m_block_cache = block_cache;
-
-	// For now we're ignoring that we've got the scratch pages
-	// allocated, full of page_block structs. Eventually hand
-	// control of that to a slab allocator.
-
-	consolidate_blocks();
-
-	// Initialize the kernel memory manager
-	addr_t end = 0;
-	for (page_block *b = m_used; b; b = b->next) {
-		if (b->virtual_address < page_offset) {
-			end = b->virtual_end();
-		} else {
-			break;
-		}
-	}
-
-	extern kutil::memory_manager g_kernel_memory_manager;
-	new (&g_kernel_memory_manager) kutil::memory_manager(
-			reinterpret_cast<void *>(end),
-			mm_grow_callback);
+	unmap_table(pml4, page_table::level::pml4, true);
 }
 
 void
 page_manager::map_offset_pointer(void **pointer, size_t length)
 {
-	addr_t *p = reinterpret_cast<addr_t *>(pointer);
-	addr_t v = *p + page_offset;
-	addr_t c = ((length - 1) / page_size) + 1;
-
-	// TODO: cleanly search/split this as a block out of used/free if possible
-	page_block *block = get_block();
-
-	// TODO: page-align
-	block->physical_address = *p;
-	block->virtual_address = v;
-	block->count = c;
-	block->flags =
-		page_block_flags::used |
-		page_block_flags::mapped |
-		page_block_flags::mmio;
-
-	m_used = page_block::insert(m_used, block);
-
-	page_table *pml4 = get_pml4();
-	page_in(pml4, *p, v, c);
-	*p = v;
+	log::debug(logs::paging, "Mapping offset pointer region at %016lx size 0x%lx", *pointer, length);
+	*pointer = kutil::offset_pointer(*pointer, page_offset);
 }
 
 void
-page_manager::dump_blocks()
+page_manager::dump_pml4(page_table *pml4, bool recurse)
 {
-	page_block::dump(m_used, "used", true);
-	page_block::dump(m_free, "free", true);
-}
-
-page_block *
-page_manager::get_block()
-{
-	page_block *block = m_block_cache;
-	if (block) {
-		m_block_cache = block->next;
-		block->next = 0;
-		return block;
-	} else {
-		kassert(0, "NYI: page_manager::get_block() needed to allocate.");
-		return nullptr;
-	}
-}
-
-void
-page_manager::free_blocks(page_block *block)
-{
-	if (!block) return;
-
-	page_block *cur = block;
-	while (cur) {
-		page_block *next = cur->next;
-		cur->zero(cur->next ? cur->next : m_block_cache);
-		cur = next;
-	}
-
-	m_block_cache = block;
+	if (pml4 == nullptr) pml4 = get_pml4();
+	pml4->dump(page_table::level::pml4, recurse);
 }
 
 page_table *
 page_manager::get_table_page()
 {
 	if (!m_page_cache) {
-		addr_t phys = 0;
-		size_t n = pop_pages(32, &phys);
-		addr_t virt = phys + page_offset;
-
-		page_block *block = get_block();
-		block->physical_address = phys;
-		block->virtual_address = virt;
-		block->count = n;
-		page_block::insert(m_used, block);
-
-		page_in(get_pml4(), phys, virt, n);
+		uintptr_t phys = 0;
+		size_t n = m_frames.allocate(32, &phys); // TODO: indicate frames must be offset-mappable
+		uintptr_t virt = phys + page_offset;
 
 		m_page_cache = reinterpret_cast<free_page_header *>(virt);
 
 		// The last one needs to be null, so do n-1
-		addr_t end = virt + (n-1) * page_size;
+		uintptr_t end = virt + (n-1) * frame_size;
 		while (virt < end) {
 			reinterpret_cast<free_page_header *>(virt)->next =
-				reinterpret_cast<free_page_header *>(virt + page_size);
-			virt += page_size;
+				reinterpret_cast<free_page_header *>(virt + frame_size);
+			virt += frame_size;
 		}
 		reinterpret_cast<free_page_header *>(virt)->next = nullptr;
 
-		log::info(logs::memory, "Mappd %d new page table pages at %lx", n, phys);
+		log::info(logs::paging, "Mappd %d new page table pages at %lx", n, phys);
 	}
 
 	free_page_header *page = m_page_cache;
 	m_page_cache = page->next;
+
 	return reinterpret_cast<page_table *>(page);
 }
 
 void
 page_manager::free_table_pages(void *pages, size_t count)
 {
-	addr_t start = reinterpret_cast<addr_t>(pages);
+	uintptr_t start = reinterpret_cast<uintptr_t>(pages);
 	for (size_t i = 0; i < count; ++i) {
-		addr_t addr = start + (i * page_size);
+		uintptr_t addr = start + (i * frame_size);
 		free_page_header *header = reinterpret_cast<free_page_header *>(addr);
 		header->count = 1;
 		header->next = m_page_cache;
@@ -330,186 +243,180 @@ page_manager::free_table_pages(void *pages, size_t count)
 	}
 }
 
-void
-page_manager::consolidate_blocks()
-{
-	m_block_cache = page_block::append(m_block_cache, page_block::consolidate(m_free));
-	m_block_cache = page_block::append(m_block_cache, page_block::consolidate(m_used));
-}
-
 void *
-page_manager::map_pages(addr_t address, size_t count)
+page_manager::map_pages(uintptr_t address, size_t count, bool user, page_table *pml4)
 {
+	if (!address) {
+		kassert(!user, "Cannot call map_pages with 0 address for user mapping");
+		address = m_addrs.allocate(count * frame_size);
+	}
+
 	void *ret = reinterpret_cast<void *>(address);
-	page_table *pml4 = get_pml4();
+	if (!pml4) pml4 = get_pml4();
 
 	while (count) {
-		kassert(m_free, "page_manager::map_pages ran out of free pages!");
+		uintptr_t phys = 0;
+		size_t n = m_frames.allocate(count, &phys);
 
-		addr_t phys = 0;
-		size_t n = pop_pages(count, &phys);
+		log::info(logs::paging, "Paging in %d pages at p:%016lx to v:%016lx into %016lx table",
+				n, phys, address, pml4);
 
-		page_block *block = get_block();
-		block->physical_address = phys;
-		block->virtual_address = address;
-		block->count = n;
-		block->flags =
-				page_block_flags::used |
-				page_block_flags::mapped;
-		page_block::insert(m_used, block);
+		page_in(pml4, phys, address, n, user);
 
-		page_in(pml4, phys, address, n);
-
-		address += n * page_size;
+		address += n * frame_size;
 		count -= n;
 	}
 
 	return ret;
 }
 
-void *
-page_manager::map_offset_pages(size_t count)
+void
+page_manager::unmap_table(page_table *table, page_table::level lvl, bool free, page_table_indices index)
 {
-	page_table *pml4 = get_pml4();
-	page_block *free = m_free;
-	page_block *prev = nullptr;
+	const int max =
+		lvl == page_table::level::pml4 ?
+			510 :
+			512;
 
-	log::debug(logs::memory, "Got request to offset map %d pages", count);
+	uintptr_t free_start = 0;
+	uintptr_t free_start_virt = 0;
+	uintptr_t free_count = 0;
 
-	while (free) {
-		if (free->count < count) {
-			prev = free;
-			free = free->next;
-			continue;
+	size_t size =
+		lvl == page_table::level::pdp ? (1<<30) :
+		lvl == page_table::level::pd  ? (1<<21) :
+		lvl == page_table::level::pt  ? (1<<12) :
+		0;
+
+	for (int i = 0; i < max; ++i) {
+		if (!table->is_present(i)) continue;
+
+		index[lvl] = i;
+
+		bool is_page =
+			lvl == page_table::level::pt ||
+			table->is_large_page(lvl, i);
+
+		if (is_page) {
+			uintptr_t frame = table->entries[i] & ~0xfffull;
+			if (!free_count || frame != free_start + free_count * size) {
+				if (free_count && free) {
+					log::debug(logs::paging,
+							"  freeing v:%016lx-%016lx p:%016lx-%016lx",
+							free_start_virt, free_start_virt + free_count * frame_size,
+							free_start, free_start + free_count * frame_size);
+
+					m_frames.free(free_start, (free_count * size) / frame_size);
+					free_count = 0;
+				}
+
+				if (!free_count) {
+					free_start = frame;
+					free_start_virt = index.addr();
+				}
+			}
+
+			free_count += 1;
+		} else {
+			page_table *next = table->get(i);
+			unmap_table(next, page_table::deeper(lvl), free, index);
 		}
-
-		page_block *used = get_block();
-		used->count = count;
-		used->physical_address = free->physical_address;
-		used->virtual_address = used->physical_address + page_offset;
-		used->flags =
-			page_block_flags::used |
-			page_block_flags::mapped;
-		page_block::insert(m_used, used);
-
-		free->physical_address += count * page_size;
-		free->count -= count;
-
-		if (free->count == 0) {
-			if (prev)
-				prev->next = free->next;
-			else
-				m_free = free->next;
-
-			free->zero(m_block_cache);
-			m_block_cache = free;
-		}
-
-		page_in(pml4, used->physical_address, used->virtual_address, count);
-		return reinterpret_cast<void *>(used->virtual_address);
 	}
 
-	return nullptr;
+	if (free_count && free) {
+		log::debug(logs::paging,
+				"  freeing v:%016lx-%016lx p:%016lx-%016lx",
+				free_start_virt, free_start_virt + free_count * frame_size,
+				free_start, free_start + free_count * frame_size);
+
+		m_frames.free(free_start, (free_count * size) / frame_size);
+	}
+	free_table_pages(table, 1);
+
+	log::debug(logs::paging, "Unmapped%s lv %d table at %016lx",
+			free ? " (and freed)" : "", lvl, table);
 }
 
 void
-page_manager::unmap_pages(void* address, size_t count)
+page_manager::unmap_pages(void* address, size_t count, page_table *pml4)
 {
-	addr_t addr = reinterpret_cast<addr_t>(address);
+	if (!pml4)
+		pml4 = get_pml4();
 
-	page_block **prev = &m_used;
-	page_block *cur = m_used;
-	while (cur && !cur->contains(addr)) {
-		prev = &cur->next;
-		cur = cur->next;
-	}
+	uintptr_t iaddr = reinterpret_cast<uintptr_t>(address);
 
-	kassert(cur, "Couldn't find existing mapped pages to unmap");
-
-	size_t size = page_size * count;
-	addr_t end = addr + size;
-
-	while (cur && cur->contains(addr)) {
-		size_t leading = addr - cur->virtual_address;
-		size_t trailing =
-			end > cur->virtual_end() ?
-			0 : (cur->virtual_end() - end);
-
-		if (leading) {
-			size_t pages = leading / page_size;
-
-			page_block *lead_block = get_block();
-			lead_block->copy(cur);
-			lead_block->next = cur;
-			lead_block->count = pages;
-
-			cur->count -= pages;
-			cur->physical_address += leading;
-			cur->virtual_address += leading;
-
-			*prev = lead_block;
-			prev = &lead_block->next;
-		}
-
-		if (trailing) {
-			size_t pages = trailing / page_size;
-
-			page_block *trail_block = get_block();
-			trail_block->copy(cur);
-			trail_block->next = cur->next;
-			trail_block->count = pages;
-			trail_block->physical_address += size;
-			trail_block->virtual_address += size;
-
-			cur->count -= pages;
-
-			cur->next = trail_block;
-		}
-
-		addr += cur->count * page_size;
-		page_block *next = cur->next;
-
-		*prev = cur->next;
-		cur->next = nullptr;
-		cur->virtual_address = 0;
-		cur->flags = cur->flags & ~(page_block_flags::used | page_block_flags::mapped);
-		m_free = page_block::insert(m_free, cur);
-
-		cur = next;
+	page_out(pml4, iaddr, count, true);
+	if (iaddr >= kernel_offset) {
+		// TODO
+		// m_addrs.free(address, count);
 	}
 }
 
+bool
+page_manager::fault_handler(uintptr_t addr)
+{
+	if (!m_addrs.contains(addr))
+		return false;
+
+	uintptr_t page = addr & ~0xfffull;
+	bool user = addr < kernel_offset;
+	map_pages(page, 1, user);
+
+	return true;
+}
+
 void
-page_manager::check_needs_page(page_table *table, unsigned index)
+page_manager::check_needs_page(page_table *table, unsigned index, bool user)
 {
 	if ((table->entries[index] & 0x1) == 1) return;
 
 	page_table *new_table = get_table_page();
 	for (int i=0; i<512; ++i) new_table->entries[i] = 0;
-	table->entries[index] = pt_to_phys(new_table) | 0xb;
+	table->entries[index] = pt_to_phys(new_table) | (user ? 0xf : 0xb);
 }
 
 void
-page_manager::page_in(page_table *pml4, addr_t phys_addr, addr_t virt_addr, size_t count)
+page_manager::page_in(page_table *pml4, uintptr_t phys_addr, uintptr_t virt_addr, size_t count, bool user, bool large)
 {
+	/*
+	log::debug(logs::paging, "page_in for table %016lx p:%016lx v:%016lx c:%4d u:%d l:%d",
+			pml4, phys_addr, virt_addr, count, user, large);
+	*/
+
 	page_table_indices idx{virt_addr};
 	page_table *tables[4] = {pml4, nullptr, nullptr, nullptr};
 
+	uint64_t flags = user ?
+		0x00f:	// writethru, user, write, present
+		0x10b;	// global, writethru, write, present
+
 	for (; idx[0] < 512; idx[0] += 1) {
-		check_needs_page(tables[0], idx[0]);
+		check_needs_page(tables[0], idx[0], user);
 		tables[1] = tables[0]->get(idx[0]);
 
 		for (; idx[1] < 512; idx[1] += 1, idx[2] = 0, idx[3] = 0) {
-			check_needs_page(tables[1], idx[1]);
+			check_needs_page(tables[1], idx[1], user);
 			tables[2] = tables[1]->get(idx[1]);
 
 			for (; idx[2] < 512; idx[2] += 1, idx[3] = 0) {
-				check_needs_page(tables[2], idx[2]);
+				if (large &&
+					idx[3] == 0 &&
+					count >= 512 &&
+					tables[2]->get(idx[2]) == nullptr) {
+					// Do a 2MiB page instead
+					tables[2]->entries[idx[2]] = phys_addr | flags | 0x80;
+					phys_addr += frame_size * 512;
+					count -= 512;
+					if (count == 0) return;
+					continue;
+				}
+
+				check_needs_page(tables[2], idx[2], user);
 				tables[3] = tables[2]->get(idx[2]);
 
 				for (; idx[3] < 512; idx[3] += 1) {
-					tables[3]->entries[idx[3]] = phys_addr | 0xb;
-					phys_addr += page_manager::page_size;
+					tables[3]->entries[idx[3]] = phys_addr | flags;
+					phys_addr += frame_size;
 					if (--count == 0) return;
 				}
 			}
@@ -520,26 +427,41 @@ page_manager::page_in(page_table *pml4, addr_t phys_addr, addr_t virt_addr, size
 }
 
 void
-page_manager::page_out(page_table *pml4, addr_t virt_addr, size_t count)
+page_manager::page_out(page_table *pml4, uintptr_t virt_addr, size_t count, bool free)
 {
 	page_table_indices idx{virt_addr};
 	page_table *tables[4] = {pml4, nullptr, nullptr, nullptr};
 
+	bool found = false;
+	uintptr_t free_start = 0;
+	unsigned free_count = 0;
+
 	for (; idx[0] < 512; idx[0] += 1) {
-		tables[1] = reinterpret_cast<page_table *>(
-				tables[0]->entries[idx[0]] & ~0xfffull);
+		tables[1] = tables[0]->get(idx[0]);
 
 		for (; idx[1] < 512; idx[1] += 1) {
-			tables[2] = reinterpret_cast<page_table *>(
-					tables[1]->entries[idx[1]] & ~0xfffull);
+			tables[2] = tables[1]->get(idx[1]);
 
 			for (; idx[2] < 512; idx[2] += 1) {
-				tables[3] = reinterpret_cast<page_table *>(
-						tables[2]->entries[idx[2]] & ~0xfffull);
+				tables[3] = tables[2]->get(idx[2]);
 
 				for (; idx[3] < 512; idx[3] += 1) {
+					uintptr_t entry = tables[3]->entries[idx[3]] & ~0xfffull;
+					if (!found || entry != free_start + free_count * frame_size) {
+						if (found && free) m_frames.free(free_start, free_count);
+						free_start = tables[3]->entries[idx[3]] & ~0xfffull;
+						free_count = 1;
+						found = true;
+					} else {
+						free_count++;
+					}
+
 					tables[3]->entries[idx[3]] = 0;
-					if (--count == 0) return;
+
+					if (--count == 0) {
+						if (free) m_frames.free(free_start, free_count);
+						return;
+					}
 				}
 			}
 		}
@@ -548,65 +470,60 @@ page_manager::page_out(page_table *pml4, addr_t virt_addr, size_t count)
 	kassert(0, "Ran to end of page_out");
 }
 
-size_t
-page_manager::pop_pages(size_t count, addr_t *address)
-{
-	kassert(m_free, "page_manager::pop_pages ran out of free pages!");
-
-	unsigned n = std::min(count, static_cast<size_t>(m_free->count));
-	*address = m_free->physical_address;
-
-	m_free->physical_address += n * page_size;
-	m_free->count -= n;
-	if (m_free->count == 0) {
-		page_block *block = m_free;
-		m_free = m_free->next;
-
-		block->zero(m_block_cache);
-		m_block_cache = block;
-	}
-
-	return n;
-}
-
-
 void
-page_table::dump(int level, uint64_t offset)
+page_table::dump(page_table::level lvl, bool recurse)
 {
-	log::info(logs::memory, "Level %d page table @ %lx (off %lx):", level, this, offset);
+	console *cons = console::get();
+
+	cons->printf("\nLevel %d page table @ %lx:\n", lvl, this);
 	for (int i=0; i<512; ++i) {
 		uint64_t ent = entries[i];
-		if (ent == 0) continue;
 
-		if ((ent & 0x1) == 0) {
-			log::info(logs::memory, "  %3d: %lx   NOT PRESENT", i, ent);
-			continue;
-		}
+		if ((ent & 0x1) == 0)
+			cons->printf("  %3d: %016lx   NOT PRESENT\n", i, ent);
 
-		if ((level == 2 || level == 3) && (ent & 0x80) == 0x80) {
-			log::info(logs::memory, "  %3d: %lx   -> Large page at     %lx",
-					i, ent, ent & ~0xfffull);
-			continue;
-		} else if (level == 1) {
-			log::info(logs::memory, "  %3d: %lx   -> Page at           %lx",
-					i, ent, ent & ~0xfffull);
-		} else {
-			log::info(logs::memory, "  %3d: %lx   -> Level %d table at %lx",
-					i, ent, level - 1, (ent & ~0xfffull) + offset);
-			continue;
-		}
+		else if ((lvl == level::pdp || lvl == level::pd) && (ent & 0x80) == 0x80)
+			cons->printf("  %3d: %016lx -> Large page at    %016lx\n", i, ent, ent & ~0xfffull);
+
+		else if (lvl == level::pt)
+			cons->printf("  %3d: %016lx -> Page at          %016lx\n", i, ent, ent & ~0xfffull);
+
+		else
+			cons->printf("  %3d: %016lx -> Level %d table at %016lx\n",
+					i, ent, deeper(lvl), (ent & ~0xfffull) + page_offset);
 	}
 
-	if (--level > 0) {
-		for (int i=0; i<512; ++i) {
-			uint64_t ent = entries[i];
-			if ((ent & 0x1) == 0) continue;
-			if ((ent & 0x80)) continue;
+	if (lvl != level::pt && recurse) {
+		for (int i=0; i<=512; ++i) {
+			if (is_large_page(lvl, i))
+				continue;
 
-			page_table *next = reinterpret_cast<page_table *>((ent & ~0xffful) + offset); 
-			next->dump(level, offset);
+			page_table *next = get(i);
+			if (next)
+				next->dump(deeper(lvl), true);
 		}
 	}
 }
 
+page_table_indices::page_table_indices(uint64_t v) :
+	index{
+		(v >> 39) & 0x1ff,
+		(v >> 30) & 0x1ff,
+		(v >> 21) & 0x1ff,
+		(v >> 12) & 0x1ff }
+{}
 
+uintptr_t
+page_table_indices::addr() const
+{
+	return
+		(index[0] << 39) |
+		(index[1] << 30) |
+		(index[2] << 21) |
+		(index[3] << 12);
+}
+
+bool operator==(const page_table_indices &l, const page_table_indices &r)
+{
+	return l[0] == r[0] && l[1] == r[1] && l[2] == r[2] && l[3] == r[3];
+}

src/kernel/page_manager.h

@@ -5,46 +5,81 @@
 #include <stddef.h>
 #include <stdint.h>
 
-#include "kutil/memory.h"
+#include "kutil/address_manager.h"
 #include "kutil/enum_bitfields.h"
+#include "kutil/linked_list.h"
+#include "kutil/slab_allocator.h"
+#include "frame_allocator.h"
+#include "kernel_memory.h"
+#include "page_table.h"
 
-struct page_block;
-struct page_table;
 struct free_page_header;
 
-
-/// Manager for allocation of physical pages.
+/// Manager for allocation and mapping of pages
 class page_manager
 {
 public:
-	/// Size of a single page.
-	static const size_t page_size = 0x1000;
+	page_manager(
+		frame_allocator &frames,
+		kutil::address_manager &addrs);
 
-	/// Start of the higher half.
-	static const addr_t high_offset = 0xffff800000000000;
+	/// Helper to get the number of pages needed for a given number of bytes.
+	/// \arg bytes  The number of bytes desired
+	/// \returns    The number of pages needed to contain the desired bytes
+	static inline size_t page_count(size_t bytes)
+	{
+		return (bytes - 1) / memory::frame_size + 1;
+	}
 
-	/// Offset from physical where page tables are mapped.
-	static const addr_t page_offset = 0xffffff8000000000;
+	/// Helper to read the PML4 table from CR3.
+	/// \returns  A pointer to the current PML4 table.
+	static inline page_table * get_pml4()
+	{
+		uintptr_t pml4 = 0;
+		__asm__ __volatile__ ( "mov %%cr3, %0" : "=r" (pml4) );
+		return reinterpret_cast<page_table *>((pml4 & ~0xfffull) + memory::page_offset);
+	}
 
-	page_manager();
+	/// Helper to set the PML4 table pointer in CR3.
+	/// \arg pml4  A pointer to the PML4 table to install.
+	static inline void set_pml4(page_table *pml4)
+	{
+		constexpr uint64_t phys_mask = ~memory::page_offset & ~0xfffull;
+		uintptr_t p = reinterpret_cast<uintptr_t>(pml4) & phys_mask;
+		__asm__ __volatile__ ( "mov %0, %%cr3" :: "r" (p) );
+	}
+
+	/// Allocate but don't switch to a new PML4 table. This table
+	/// should only have global kernel pages mapped.
+	/// \returns      A pointer to the PML4 table
+	page_table * create_process_map();
+
+	/// Deallocate a process' PML4 table and entries.
+	/// \arg pml4  The process' PML4 table
+	void delete_process_map(page_table *pml4);
+
+	/// Copy a process' memory mappings (and memory pages).
+	/// \arg from  Page table to copy from
+	/// \arg lvl   Level of the given tables (default is PML4)
+	/// \returns   The new page table
+	page_table * copy_table(page_table *from,
+			page_table::level lvl = page_table::level::pml4,
+			page_table_indices index = {});
 
 	/// Allocate and map pages into virtual memory.
-	/// \arg address  The virtual address at which to map the pages
+	/// \arg address  The virtual address at which to map the pages, or zero
+	///               for any free kernel space.
 	/// \arg count    The number of pages to map
+	/// \arg user     True is this memory is user-accessible
+	/// \arg pml4     The pml4 to map into - null for the current one
 	/// \returns      A pointer to the start of the mapped region
-	void * map_pages(addr_t address, size_t count);
+	void * map_pages(uintptr_t address, size_t count, bool user = false, page_table *pml4 = nullptr);
 
-	/// Allocate and map contiguous pages into virtual memory, with
-	/// a constant offset from their physical address.
-	/// \arg count    The number of pages to map
-	/// \returns      A pointer to the start of the mapped region, or
-	/// nullptr if no region could be found to fit the request.
-	void * map_offset_pages(size_t count);
-
-	/// Unmap existing pages from memory.
+	/// Unmap and free existing pages from memory.
 	/// \arg address  The virtual address of the memory to unmap
 	/// \arg count    The number of pages to unmap
-	void unmap_pages(void *address, size_t count);
+	/// \arg pml4     The pml4 to unmap from - null for the current one
+	void unmap_pages(void *address, size_t count, page_table *pml4 = nullptr);
 
 	/// Offset-map a pointer. No physical pages will be mapped.
 	/// \arg pointer  Pointer to a pointer to the memory area to be mapped
@@ -54,43 +89,41 @@ public:
 	/// Get the physical address of an offset-mapped pointer
 	/// \arg p   Virtual address of memory that has been offset-mapped
 	/// \returns Physical address of the memory pointed to by p
-	inline addr_t offset_phys(void *p) const
+	inline uintptr_t offset_phys(void *p) const
 	{
-		return reinterpret_cast<addr_t>(kutil::offset_pointer(p, -page_offset));
+		return reinterpret_cast<uintptr_t>(kutil::offset_pointer(p, -memory::page_offset));
 	}
 
 	/// Get the virtual address of an offset-mapped physical address
 	/// \arg a   Physical address of memory that has been offset-mapped
 	/// \returns Virtual address of the memory at address a
-	inline void * offset_virt(addr_t a) const
+	inline void * offset_virt(uintptr_t a) const
 	{
-		return kutil::offset_pointer(reinterpret_cast<void *>(a), page_offset);
+		return kutil::offset_pointer(reinterpret_cast<void *>(a), memory::page_offset);
 	}
 
-	/// Log the current free/used block lists.
-	void dump_blocks();
+	/// Dump the given or current PML4 to the console
+	/// \arg pml4     The page table to use, null for the current one
+	/// \arg recurse  Whether to print sub-tables
+	void dump_pml4(page_table *pml4 = nullptr, bool recurse = true);
 
 	/// Get the system page manager.
 	/// \returns  A pointer to the system page manager
 	static page_manager * get();
 
+	/// Get a pointer to the kernel's PML4
+	inline page_table * get_kernel_pml4() { return m_kernel_pml4; }
+
+	/// Attempt to handle a page fault.
+	/// \arg addr  Address that triggered the fault
+	/// \returns   True if the fault was handled
+	bool fault_handler(uintptr_t addr);
+
 private:
-	/// Set up the memory manager from bootstraped memory
-	void init(
-			page_block *free,
-			page_block *used,
-			page_block *block_cache);
-
-	/// Initialize the virtual memory manager based on this object's state
-	void init_memory_manager();
-
-	/// Create a `page_block` struct or pull one from the cache.
-	/// \returns  An empty `page_block` struct
-	page_block * get_block();
-
-	/// Return a list of `page_block` structs to the cache.
-	/// \arg block   A list of `page_block` structs
-	void free_blocks(page_block *block);
+	/// Copy a physical page
+	/// \arg orig  Physical address of the page to copy
+	/// \returns   Physical address of the new page
+	uintptr_t copy_page(uintptr_t orig);
 
 	/// Allocate a page for a page table, or pull one from the cache
 	/// \returns  An empty page mapped in page space
@@ -101,69 +134,51 @@ private:
 	/// \arg count  Number of pages in the range
 	void free_table_pages(void *pages, size_t count);
 
-	/// Consolidate the free and used block lists. Return freed blocks
-	/// to the cache.
-	void consolidate_blocks();
-
-	/// Helper to read the PML4 table from CR3.
-	/// \returns  A pointer to the current PML4 table.
-	static inline page_table * get_pml4()
-	{
-		addr_t pml4 = 0;
-		__asm__ __volatile__ ( "mov %%cr3, %0" : "=r" (pml4) );
-		return reinterpret_cast<page_table *>((pml4 & ~0xfffull) + page_offset);
-	}
-
-	/// Helper to set the PML4 table pointer in CR3.
-	/// \arg pml4  A pointer to the PML4 table to install.
-	static inline void set_pml4(page_table *pml4)
-	{
-		addr_t p = reinterpret_cast<addr_t>(pml4) - page_offset;
-		__asm__ __volatile__ ( "mov %0, %%cr3" :: "r" (p & ~0xfffull) );
-	}
-
 	/// Helper function to allocate a new page table. If table entry `i` in
 	/// table `base` is empty, allocate a new page table and point `base[i]` at
 	/// it.
 	/// \arg base  Existing page table being indexed into
 	/// \arg i     Index into the existing table to check
-	void check_needs_page(page_table *base, unsigned i);
+	/// \art user  True if this is a userspace mapping
+	void check_needs_page(page_table *base, unsigned i, bool user);
 
 	/// Low-level routine for mapping a number of pages into the given page table.
 	/// \arg pml4       The root page table to map into
 	/// \arg phys_addr  The starting physical address of the pages to be mapped
 	/// \arg virt_addr  The starting virtual address ot the memory to be mapped
 	/// \arg count      The number of pages to map
+	/// \arg user       True if this is a userspace mapping
+	/// \arg large      Whether to allow large pages
 	void page_in(
 			page_table *pml4,
-			addr_t phys_addr,
-			addr_t virt_addr,
-			size_t count);
+			uintptr_t phys_addr,
+			uintptr_t virt_addr,
+			size_t count,
+			bool user = false,
+			bool large = false);
 
 	/// Low-level routine for unmapping a number of pages from the given page table.
 	/// \arg pml4       The root page table for this mapping
 	/// \arg virt_addr  The starting virtual address ot the memory to be unmapped
 	/// \arg count      The number of pages to unmap
+	/// \arg free       Whether to return the pages to the frame allocator
 	void page_out(
 			page_table *pml4,
-			addr_t virt_addr,
-			size_t count);
+			uintptr_t virt_addr,
+			size_t count,
+			bool free = false);
 
-	/// Get free pages from the free list. Only pages from the first free block
-	/// are returned, so the number may be less than requested, but they will
-	/// be contiguous. Pages will not be mapped into virtual memory.
-	/// \arg count    The maximum number of pages to get
-	/// \arg address  [out] The address of the first page
-	/// \returns      The number of pages retrieved
-	size_t pop_pages(size_t count, addr_t *address);
+	/// Low-level routine for unmapping an entire table of memory at once
+	void unmap_table(page_table *table, page_table::level lvl, bool free,
+			page_table_indices index = {});
 
-	page_block *m_free; ///< Free pages list
-	page_block *m_used; ///< In-use pages list
-
-	page_block *m_block_cache; ///< Cache of unused page_block structs
+	page_table *m_kernel_pml4; ///< The PML4 of just kernel pages
 	free_page_header *m_page_cache; ///< Cache of free pages to use for tables
 
-	friend void memory_initialize(const void *, size_t, size_t);
+	frame_allocator &m_frames;
+	kutil::address_manager &m_addrs;
+
+	friend class memory_bootstrap;
 	page_manager(const page_manager &) = delete;
 };
 
@@ -172,132 +187,6 @@ extern page_manager g_page_manager;
 
 inline page_manager * page_manager::get() { return &g_page_manager; }
 
-/// Flags used by `page_block`.
-enum class page_block_flags : uint32_t
-{
-	free         = 0x00000000,  ///< Not a flag, value for free memory
-	used         = 0x00000001,  ///< Memory is in use
-	mapped       = 0x00000002,  ///< Memory is mapped to virtual address
-
-	mmio         = 0x00000010,  ///< Memory is a MMIO region
-	nonvolatile  = 0x00000020,  ///< Memory is non-volatile storage
-
-	pending_free = 0x10000000,  ///< Memory should be freed
-	acpi_wait    = 0x40000000,  ///< Memory should be freed after ACPI init
-	permanent    = 0x80000000,  ///< Memory is permanently unusable
-
-	max_flags
-};
-IS_BITFIELD(page_block_flags);
-
-
-/// A block of contiguous pages. Each `page_block` represents contiguous
-/// physical pages with the same attributes. A `page_block *` is also a
-/// linked list of such structures.
-struct page_block
-{
-	addr_t physical_address;
-	addr_t virtual_address;
-	uint32_t count;
-	page_block_flags flags;
-	page_block *next;
-
-	inline bool has_flag(page_block_flags f) const { return bitfield_has(flags, f); }
-	inline addr_t physical_end() const { return physical_address + (count * page_manager::page_size); }
-	inline addr_t virtual_end() const { return virtual_address + (count * page_manager::page_size); }
-
-	inline bool contains(addr_t vaddr) const { return vaddr >= virtual_address && vaddr < virtual_end(); }
-	inline bool contains_physical(addr_t addr) const { return addr >= physical_address && addr < physical_end(); }
-
-	/// Helper to zero out a block and optionally set the next pointer.
-	/// \arg next  [optional] The value for the `next` pointer
-	void zero(page_block *set_next = nullptr);
-
-	/// Helper to copy a bock from another block
-	/// \arg other  The block to copy from
-	void copy(page_block *other);
-
-	/// \name Page block linked list functions
-	/// Functions to act on a `page_block *` as a linked list
-	/// @{
-
-	/// Count the items in the given linked list.
-	/// \arg list  The list to count
-	/// \returns   The number of entries in the list.
-	static size_t length(page_block *list);
-
-	/// Append a block or list to the given list.
-	/// \arg list   The list to append to
-	/// \arg extra  The list or block to be appended 
-	/// \returns    The new list head
-	static page_block * append(page_block *list, page_block *extra);
-
-	/// Sorted-insert of a block into the list by address.
-	/// \arg list   The list to insert into
-	/// \arg block  The single block to insert
-	/// \returns    The new list head
-	static page_block * insert(page_block *list, page_block *block);
-
-	/// Compare two blocks by address.
-	/// \arg lhs   The left-hand comparator
-	/// \arg rhs   The right-hand comparator
-	/// \returns   <0 if lhs is sorts earlier, >0 if lhs sorts later, 0 for equal
-	static int compare(const page_block *lhs, const page_block *rhs);
-
-	/// Traverse the list, joining adjacent blocks where possible.
-	/// \arg list  The list to consolidate
-	/// \returns   A linked list of freed page_block structures.
-	static page_block * consolidate(page_block *list);
-
-	/// Traverse the list, printing debug info on this list.
-	/// \arg list  The list to print
-	/// \arg name  [optional] String to print as the name of this list
-	/// \arg show_permanent [optional] If false, hide unmapped blocks
-	static void dump(page_block *list, const char *name = nullptr, bool show_unmapped = false);
-
-	/// @}
-
-};
-
-
-
-/// Struct to allow easy accessing of a memory page being used as a page table.
-struct page_table
-{
-	using pm = page_manager;
-
-	uint64_t entries[512];
-
-	inline page_table * get(int i) const {
-		uint64_t entry = entries[i];
-		if ((entry & 0x1) == 0) return nullptr;
-		return reinterpret_cast<page_table *>((entry & ~0xfffull) + pm::page_offset);
-	}
-
-	inline void set(int i, page_table *p, uint16_t flags) {
-		entries[i] = (reinterpret_cast<uint64_t>(p) - pm::page_offset) | (flags & 0xfff);
-	}
-
-	void dump(int level = 4, uint64_t offset = page_manager::page_offset);
-};
-
-
-/// Helper struct for computing page table indices of a given address.
-struct page_table_indices
-{
-	page_table_indices(uint64_t v = 0) :
-		index{
-			(v >> 39) & 0x1ff,
-			(v >> 30) & 0x1ff,
-			(v >> 21) & 0x1ff,
-			(v >> 12) & 0x1ff }
-	{}
-
-	/// Get the index for a given level of page table.
-	uint64_t & operator[](size_t i) { return index[i]; }
-	uint64_t index[4]; ///< Indices for each level of tables.
-};
-
 
 /// Calculate a page-aligned address.
 /// \arg p    The address to align.
@@ -306,8 +195,8 @@ template <typename T> inline T
 page_align(T p)
 {
 	return reinterpret_cast<T>(
-		((reinterpret_cast<addr_t>(p) - 1) & ~(page_manager::page_size - 1))
-		+ page_manager::page_size);
+		((reinterpret_cast<uintptr_t>(p) - 1) & ~(memory::frame_size - 1))
+		+ memory::frame_size);
 }
 
 /// Calculate a page-table-aligned address. That is, an address that is
@@ -321,11 +210,9 @@ page_table_align(T p)
 }
 
 
-/// Calculate the number of pages needed for the give number of bytes.
-/// \arg n   Number of bytes
-/// \returns Number of pages
-inline size_t page_count(size_t n) { return ((n - 1) / page_manager::page_size) + 1; }
-
-
 /// Bootstrap the memory managers.
-void memory_initialize(const void *memory_map, size_t map_length, size_t desc_length);
+kutil::allocator & memory_initialize(
+		uint16_t scratch_pages,
+		const void *memory_map,
+		size_t map_length,
+		size_t desc_length);

src/kernel/page_table.h

@@ -0,0 +1,63 @@
+#pragma once
+/// \file page_table.h
+/// Helper structures for dealing with page tables.
+
+#include <stdint.h>
+#include "kernel_memory.h"
+
+class page_manager;
+
+/// Struct to allow easy accessing of a memory page being used as a page table.
+struct page_table
+{
+	enum class level : unsigned { pml4, pdp, pd, pt };
+	inline static level deeper(level l) {
+		return static_cast<level>(static_cast<unsigned>(l) + 1);
+	}
+
+	uint64_t entries[512];
+
+	inline page_table * get(int i, uint16_t *flags = nullptr) const {
+		uint64_t entry = entries[i];
+		if ((entry & 0x1) == 0) return nullptr;
+		if (flags) *flags = entry & 0xfffull;
+		return reinterpret_cast<page_table *>((entry & ~0xfffull) + memory::page_offset);
+	}
+
+	inline void set(int i, page_table *p, uint16_t flags) {
+		entries[i] = (reinterpret_cast<uint64_t>(p) - memory::page_offset) | (flags & 0xfff);
+	}
+
+	inline bool is_present(int i) const { return (entries[i] & 0x1) == 0x1; }
+
+	inline bool is_large_page(level l, int i) const {
+		return
+			(l == level::pdp || l == level::pd) &&
+			(entries[i] & 0x80) == 0x80;
+	}
+
+	void dump(
+		level lvl = level::pml4,
+		bool recurse = true);
+};
+
+
+/// Helper struct for computing page table indices of a given address.
+struct page_table_indices
+{
+	page_table_indices(uint64_t v = 0);
+
+	uintptr_t addr() const;
+
+	inline operator uintptr_t() const { return addr(); }
+
+	/// Get the index for a given level of page table.
+	uint64_t & operator[](int i) { return index[i]; }
+	uint64_t operator[](int i) const { return index[i]; }
+	uint64_t & operator[](page_table::level i) { return index[static_cast<unsigned>(i)]; }
+	uint64_t operator[](page_table::level i) const { return index[static_cast<unsigned>(i)]; }
+	uint64_t index[4]; ///< Indices for each level of tables.
+};
+
+bool operator==(const page_table_indices &l, const page_table_indices &r);
+

src/kernel/pci.cpp

@@ -1,11 +1,32 @@
 #include "kutil/assert.h"
+#include "console.h"
 #include "log.h"
 #include "interrupts.h"
 #include "pci.h"
 
-struct pci_cap_msi :
-	public pci_cap
+struct pci_cap_msi
 {
+	pci_cap::type id;
+	uint8_t next;
+	uint16_t control;
+} __attribute__ ((packed));
+
+struct pci_cap_msi32
+{
+	pci_cap::type id;
+	uint8_t next;
+	uint16_t control;
+	uint32_t address;
+	uint16_t data;
+	uint16_t reserved;
+	uint32_t mask;
+	uint32_t pending;
+} __attribute__ ((packed));
+
+struct pci_cap_msi64
+{
+	pci_cap::type id;
+	uint8_t next;
 	uint16_t control;
 	uint64_t address;
 	uint16_t data;
@@ -14,17 +35,33 @@ struct pci_cap_msi :
 	uint32_t pending;
 } __attribute__ ((packed));
 
-struct pci_cap_msix :
-	public pci_cap
-{
-	uint16_t control;
-	uint64_t address;
-	uint16_t data;
-	uint16_t reserved;
-	uint32_t mask;
-	uint32_t pending;
-} __attribute__ ((packed));
 
+void dump_msi(pci_cap_msi *cap)
+{
+	auto cons = console::get();
+	cons->printf("MSI Cap:\n");
+	cons->printf("     id: %02x\n", cap->id);
+	cons->printf("   next: %02x\n", cap->next);
+	cons->printf("control: %04x\n", cap->control);
+	if (cap->control & 0x0080) {
+		pci_cap_msi64 *cap64 = reinterpret_cast<pci_cap_msi64 *>(cap);
+		cons->printf("address: %016x\n", cap64->address);
+		cons->printf("   data: %04x\n", cap64->data);
+		if (cap->control & 0x100) {
+			cons->printf("   mask: %08x\n", cap64->mask);
+			cons->printf("pending: %08x\n", cap64->pending);
+		}
+	} else {
+		pci_cap_msi32 *cap32 = reinterpret_cast<pci_cap_msi32 *>(cap);
+		cons->printf("address: %08x\n", cap32->address);
+		cons->printf("   data: %04x\n", cap32->data);
+		if (cap->control & 0x100) {
+			cons->printf("   mask: %08x\n", cap32->mask);
+			cons->printf("pending: %08x\n", cap32->pending);
+		}
+	}
+	cons->putc('\n');
+};
 
 pci_device::pci_device() :
 	m_base(nullptr),
@@ -60,19 +97,25 @@ pci_device::pci_device(pci_group &group, uint8_t bus, uint8_t device, uint8_t fu
 	uint16_t *command = reinterpret_cast<uint16_t *>(&m_base[1]);
 	*command |= 0x400; // Mask old INTx style interrupts
 
-	log::info(logs::device, "Found PCIe device at %02d:%02d:%d of type %d.%d id %04x:%04x",
-			bus, device, func, m_class, m_subclass, m_vendor, m_device);
+	uint16_t *status = command + 1;
 
-	// Walk the extended capabilities list
-	uint8_t next = m_base[13] & 0xff;
-	while (next) {
-		pci_cap *cap = reinterpret_cast<pci_cap *>(kutil::offset_pointer(m_base, next));
-		next = cap->next;
+	log::info(logs::device, "Found PCIe device at %02d:%02d:%d of type %x.%x.%x id %04x:%04x",
+			bus, device, func, m_class, m_subclass, m_progif, m_vendor, m_device);
 
-		if (cap->id == pci_cap::type::msi) {
-			m_msi = cap;
-			pci_cap_msi *mcap = reinterpret_cast<pci_cap_msi *>(cap);
-			mcap->control |= ~0x1; // Mask interrupts
+	if (*status & 0x0010) {
+		// Walk the extended capabilities list
+		uint8_t next = m_base[13] & 0xff;
+		while (next) {
+			pci_cap *cap = reinterpret_cast<pci_cap *>(kutil::offset_pointer(m_base, next));
+			next = cap->next;
+			log::debug(logs::device, "  - found PCI cap type %02x", cap->id);
+
+			if (cap->id == pci_cap::type::msi) {
+				m_msi = cap;
+				pci_cap_msi *mcap = reinterpret_cast<pci_cap_msi *>(cap);
+				mcap->control &= ~0x70; // at most 1 vector allocated
+				mcap->control |= 0x01; // Enable interrupts, at most 1 vector allocated
+			}
 		}
 	}
 }
@@ -106,14 +149,24 @@ pci_device::set_bar(unsigned i, uint32_t val)
 }
 
 void
-pci_device::write_msi_regs(addr_t address, uint16_t data)
+pci_device::write_msi_regs(uintptr_t address, uint16_t data)
 {
 	kassert(m_msi, "Tried to write MSI for a device without that cap");
 	if (m_msi->id == pci_cap::type::msi) {
 		pci_cap_msi *mcap = reinterpret_cast<pci_cap_msi *>(m_msi);
-		mcap->address = address;
-		mcap->data = data;
-		mcap->control |= 1;
+		if (mcap->control & 0x0080) {
+			pci_cap_msi64 *mcap64 = reinterpret_cast<pci_cap_msi64 *>(m_msi);
+			mcap64->address = address;
+			mcap64->data = data;
+		} else {
+			pci_cap_msi32 *mcap32 = reinterpret_cast<pci_cap_msi32 *>(m_msi);
+			mcap32->address = address;
+			mcap32->data = data;
+		}
+		uint16_t control = mcap->control;
+		control &= 0xff8f; // We're allocating one vector, clear 6::4
+		control |= 0x0001; // Enable MSI
+		mcap->control = control;
 	} else {
 		kassert(0, "MIS-X is NYI");
 	}

src/kernel/pci.h

@@ -1,6 +1,7 @@
 #pragma once
 /// \file pci.h
 /// PCI devices and groups
+
 #include <stdint.h>
 #include "kutil/memory.h"
 
@@ -75,7 +76,7 @@ public:
 	/// Write to the MSI registers
 	/// \arg addr  The address to write to the MSI address registers
 	/// \arg data  The value to write to the MSI data register
-	void write_msi_regs(addr_t addr, uint16_t data);
+	void write_msi_regs(uintptr_t addr, uint16_t data);
 
 	/// Get a bus address, given the bus/device/function numbers.
 	/// \arg bus    Number of the bus

src/kernel/process.cpp

@@ -0,0 +1,228 @@
+#include "kutil/heap_allocator.h"
+#include "cpu.h"
+#include "debug.h"
+#include "log.h"
+#include "process.h"
+#include "scheduler.h"
+
+extern "C" void task_fork_return_thunk();
+extern kutil::heap_allocator g_kernel_heap; // TODO: this is a bad hack to get access to the heap
+
+void
+process::exit(uint32_t code)
+{
+	return_code = code;
+	flags -= process_flags::running;
+	page_manager::get()->delete_process_map(pml4);
+}
+
+pid_t
+process::fork()
+{
+	auto &sched = scheduler::get();
+	auto *child = sched.create_process();
+	kassert(child, "process::fork() got null child");
+
+	child->ppid = pid;
+	child->flags =
+		process_flags::running |
+		process_flags::ready;
+
+	sched.m_runlists[child->priority].push_back(child);
+
+	child->pml4 = page_manager::get()->copy_table(pml4);
+	kassert(child->pml4, "process::fork() got null pml4");
+
+	child->rsp3 = bsp_cpu_data.rsp3;
+	child->setup_kernel_stack();
+
+	log::debug(logs::task, "Copied process %d to %d",
+			pid, child->pid);
+
+	log::debug(logs::task, "    PML4 %016lx", child->pml4);
+	log::debug(logs::task, "    RSP3 %016lx", child->rsp3);
+	log::debug(logs::task, "    RSP0 %016lx", child->rsp0);
+
+	// Initialize a new empty stack with a fake saved state
+	// for returning out of syscall_handler_prelude
+	size_t ret_seg_size = sizeof(uintptr_t) * 8;
+	child->rsp -= ret_seg_size;
+
+	void *this_ret_seg =
+		reinterpret_cast<void*>(rsp0 - ret_seg_size);
+	void *child_ret_seg =
+		reinterpret_cast<void*>(child->rsp);
+	kutil::memcpy(child_ret_seg, this_ret_seg, ret_seg_size);
+
+	child->add_fake_task_return(
+		reinterpret_cast<uintptr_t>(task_fork_return_thunk));
+
+	log::debug(logs::task, "    RSP  %016lx", child->rsp);
+
+	return child->pid;
+}
+
+void *
+process::setup_kernel_stack()
+{
+	constexpr unsigned null_frame_entries = 2;
+	constexpr size_t null_frame_size = null_frame_entries * sizeof(uint64_t);
+
+	void *stack_bottom = g_kernel_heap.allocate(initial_stack_size);
+	kutil::memset(stack_bottom, 0, initial_stack_size);
+
+	log::debug(logs::memory, "Created kernel stack at %016lx size 0x%lx",
+			stack_bottom, initial_stack_size);
+
+	void *stack_top =
+		kutil::offset_pointer(stack_bottom,
+				initial_stack_size - null_frame_size);
+
+	uint64_t *null_frame = reinterpret_cast<uint64_t*>(stack_top);
+	for (unsigned i = 0; i < null_frame_entries; ++i)
+		null_frame[i] = 0;
+
+	kernel_stack_size = initial_stack_size;
+	kernel_stack = reinterpret_cast<uintptr_t>(stack_bottom);
+	rsp0 = reinterpret_cast<uintptr_t>(stack_top);
+	rsp = rsp0;
+
+	return stack_top;
+}
+
+void
+process::add_fake_task_return(uintptr_t rip)
+{
+	rsp -= sizeof(uintptr_t) * 7;
+	uintptr_t *stack = reinterpret_cast<uintptr_t*>(rsp);
+
+	stack[6] = rip;        // return rip
+	stack[5] = rsp0;       // rbp
+	stack[4] = 0xbbbbbbbb; // rbx
+	stack[3] = 0x12121212; // r12
+	stack[2] = 0x13131313; // r13
+	stack[1] = 0x14141414; // r14
+	stack[0] = 0x15151515; // r15
+}
+
+bool
+process::wait_on_signal(uint64_t sigmask)
+{
+	waiting = process_wait::signal;
+	waiting_info = sigmask;
+	flags -= process_flags::ready;
+	return true;
+}
+
+bool
+process::wait_on_child(uint32_t pid)
+{
+	waiting = process_wait::child;
+	waiting_info = pid;
+	flags -= process_flags::ready;
+	return true;
+}
+
+bool
+process::wait_on_time(uint64_t time)
+{
+	waiting = process_wait::time;
+	waiting_info = time;
+	flags -= process_flags::ready;
+	return true;
+}
+
+bool
+process::wait_on_send(uint32_t target_id)
+{
+	scheduler &s = scheduler::get();
+	process *target = s.get_process_by_id(target_id);
+	if (!target) return false;
+
+	if (!target->wake_on_receive(this)) {
+		waiting = process_wait::send;
+		waiting_info = target_id;
+		flags -= process_flags::ready;
+	}
+
+	return true;
+}
+
+bool
+process::wait_on_receive(uint32_t source_id)
+{
+	scheduler &s = scheduler::get();
+	process *source = s.get_process_by_id(source_id);
+	if (!source) return false;
+
+	if (!source->wake_on_send(this)) {
+		waiting = process_wait::receive;
+		waiting_info = source_id;
+		flags -= process_flags::ready;
+		return true;
+	}
+
+	return false;
+}
+
+bool
+process::wake_on_signal(int signal)
+{
+	if (waiting != process_wait::signal ||
+		(waiting_info & (1 << signal)) == 0)
+		return false;
+
+	waiting = process_wait::none;
+	flags += process_flags::ready;
+	return true;
+}
+
+bool
+process::wake_on_child(process *child)
+{
+	if (waiting != process_wait::child ||
+		(waiting_info && waiting_info != child->pid))
+		return false;
+
+	waiting = process_wait::none;
+	flags += process_flags::ready;
+	return true;
+}
+
+bool
+process::wake_on_time(uint64_t now)
+{
+	if (waiting != process_wait::time ||
+		waiting_info > now)
+		return false;
+
+	waiting = process_wait::none;
+	flags += process_flags::ready;
+	return true;
+}
+
+bool
+process::wake_on_send(process *target)
+{
+	if (waiting != process_wait::send ||
+		waiting_info != target->pid)
+		return false;
+
+	waiting = process_wait::none;
+	flags += process_flags::ready;
+	return true;
+}
+
+
+bool
+process::wake_on_receive(process *source)
+{
+	if (waiting != process_wait::receive ||
+		waiting_info != source->pid)
+		return false;
+
+	waiting = process_wait::none;
+	flags += process_flags::ready;
+	return true;
+}
+

src/kernel/process.h

@@ -0,0 +1,145 @@
+#pragma once
+/// \file process.h
+/// The processes and related definitions
+
+#include <stdint.h>
+#include "kutil/enum_bitfields.h"
+#include "kutil/linked_list.h"
+#include "page_manager.h"
+
+typedef int32_t pid_t;
+struct cpu_state;
+
+
+enum class process_flags : uint32_t
+{
+	running   = 0x00000001,
+	ready     = 0x00000002,
+	loading   = 0x00000004,
+
+	const_pri = 0x80000000,
+
+	none      = 0x00000000
+};
+IS_BITFIELD(process_flags);
+
+enum class process_wait : uint8_t
+{
+	none,
+	signal,
+	child,
+	time,
+	send,
+	receive
+};
+
+/// A process.
+/// 
+struct process
+{
+	static const size_t initial_stack_size = 0x1000;
+
+	// Fields used by assembly routines go first.  If you change any of these,
+	// be sure to change the assembly definitions in 'tasking.inc'
+	uintptr_t rsp;
+	uintptr_t rsp0;
+	uintptr_t rsp3;
+	page_table *pml4;
+	// End of assembly fields
+
+	pid_t pid;
+	pid_t ppid;
+
+	process_flags flags;
+
+	uint16_t quanta;
+
+	uint8_t priority;
+
+	process_wait waiting;
+	uint64_t waiting_info;
+
+	uint32_t return_code;
+
+	uint32_t reserved1;
+
+	uintptr_t kernel_stack;
+	size_t kernel_stack_size;
+
+	/// Terminate this process.
+	/// \arg code   The return code to exit with.
+	void exit(unsigned code);
+
+	/// Copy this process.
+	/// \returns    Returns the child's pid to the parent, and
+	///             0 to the child.
+	pid_t fork();
+
+	/// Unready this process until it gets a signal
+	/// \arg sigmask  A bitfield of signals to wake on
+	/// \returns  Whether the process should be rescheduled
+	bool wait_on_signal(uint64_t sigmask);
+
+	/// Unready this process until a child exits
+	/// \arg pid  PID of the child to wait for, or 0 for any
+	/// \returns  Whether the process should be rescheduled
+	bool wait_on_child(uint32_t pid);
+
+	/// Unready this process until after the given time
+	/// \arg time  The time after which to wake
+	/// \returns  Whether the process should be rescheduled
+	bool wait_on_time(uint64_t time);
+
+	/// Try to send to the target process, becoming unready if it
+	/// is not waiting on receive.
+	/// \arg target_id  The process to send to
+	/// \returns  Whether the process should be rescheduled
+	bool wait_on_send(uint32_t target_id);
+
+	/// Try to receive from one or more processes, becoming unready
+	/// if none of them are waiting on a send to this process.
+	/// \arg source_id  The process to receive from
+	/// \returns  Whether the process should be rescheduled
+	bool wait_on_receive(uint32_t source_id);
+
+	/// If this process is waiting on the given signal, wake it
+	/// \argument signal  The signal sent to the process
+	/// \returns  True if this wake was handled
+	bool wake_on_signal(int signal);
+
+	/// If this process is waiting on the given child, wake it
+	/// \argument child  The process that exited
+	/// \returns  True if this wake was handled
+	bool wake_on_child(process *child);
+
+	/// If this process is waiting on a time, check it
+	/// \argument now  The current time
+	/// \returns  True if this wake was handled
+	bool wake_on_time(uint64_t now);
+
+	/// If this process is waiting to send to this target, wake it
+	/// \argument target  The target process
+	/// \returns  True if this wake was handled
+	bool wake_on_send(process *target);
+
+	/// If this process is waiting to receieve from this source, wake it
+	/// \argument source  The process that is sending
+	/// \returns  True if this wake was handled
+	bool wake_on_receive(process *source);
+
+private:
+	friend class scheduler;
+
+	/// Set up a new empty kernel stack for this process. Sets rsp0 on this
+	/// process object, but also returns it.
+	/// \returns   The new rsp0 as a pointer
+	void * setup_kernel_stack();
+
+	/// Initialize this process' kenrel stack with a fake return segment for
+	/// returning out of task_switch.
+	/// \arg rip  The rip to return to
+	void add_fake_task_return(uintptr_t rip);
+};
+
+using process_list = kutil::linked_list<process>;
+using process_node = process_list::item_type;

src/kernel/push_all.inc

@@ -0,0 +1,82 @@
+struc REGS
+.r15           resq 1  ; 0x00
+.r14           resq 1  ; 0x08
+.r13           resq 1  ; 0x10
+.r12           resq 1  ; 0x18
+.r11           resq 1  ; 0x20
+.r10           resq 1  ; 0x28
+.r9            resq 1  ; 0x30
+.r8            resq 1  ; 0x38
+
+.rdi           resq 1  ; 0x40
+.rsi           resq 1  ; 0x48
+.rbp           resq 1  ; 0x50
+.rbx           resq 1  ; 0x58
+.rdx           resq 1  ; 0x60
+.rcx           resq 1  ; 0x68
+.rax           resq 1  ; 0x70
+
+.int           resq 1  ; 0x78
+.err           resq 1  ; 0x80
+.rip           resq 1  ; 0x88
+.cs3           resq 1  ; 0x90
+.rflags        resq 1  ; 0x98
+.rsp3          resq 1  ; 0xa0
+.ss3           resq 1  ; 0xa8
+endstruc
+
+regs_total_size equ 0xb0
+regs_extra_size equ 0x78
+
+%macro push_all 0
+	sub rsp, regs_extra_size
+
+	mov [rsp + REGS.rax], rax
+	mov [rsp + REGS.rcx], rcx
+	mov [rsp + REGS.rdx], rdx
+	mov [rsp + REGS.rbx], rbx
+	mov [rsp + REGS.rbp], rbp
+	mov [rsp + REGS.rsi], rsi
+	mov [rsp + REGS.rdi], rdi
+
+	mov [rsp + REGS.r8 ], r8
+	mov [rsp + REGS.r9 ], r9
+	mov [rsp + REGS.r10], r10
+	mov [rsp + REGS.r11], r11
+	mov [rsp + REGS.r12], r12
+	mov [rsp + REGS.r13], r13
+	mov [rsp + REGS.r14], r14
+	mov [rsp + REGS.r15], r15
+%endmacro
+
+%macro pop_all 0
+	mov rax, [rsp + REGS.rax]
+	mov rcx, [rsp + REGS.rcx]
+	mov rdx, [rsp + REGS.rdx]
+	mov rbx, [rsp + REGS.rbx]
+	mov rbp, [rsp + REGS.rbp]
+	mov rsi, [rsp + REGS.rsi]
+	mov rdi, [rsp + REGS.rdi]
+
+	mov r8,  [rsp + REGS.r8 ]
+	mov r9,  [rsp + REGS.r9 ]
+	mov r10, [rsp + REGS.r10]
+	mov r11, [rsp + REGS.r11]
+	mov r12, [rsp + REGS.r12]
+	mov r13, [rsp + REGS.r13]
+	mov r14, [rsp + REGS.r14]
+	mov r15, [rsp + REGS.r15]
+
+	add rsp, regs_extra_size
+%endmacro
+
+%macro check_swap_gs 0
+	mov rax, [rsp+0x90]
+	and rax, 0x03 ; mask out the RPL
+	cmp rax, 0x03
+	jne %%noswapgs
+	swapgs
+%%noswapgs:
+%endmacro
+
+; vim: ft=asm

src/kernel/scheduler.cpp

@@ -0,0 +1,325 @@
+#include "apic.h"
+#include "console.h"
+#include "cpu.h"
+#include "debug.h"
+#include "gdt.h"
+#include "interrupts.h"
+#include "io.h"
+#include "kernel_memory.h"
+#include "log.h"
+#include "msr.h"
+#include "page_manager.h"
+#include "scheduler.h"
+
+#include "elf/elf.h"
+#include "kutil/assert.h"
+
+using memory::initial_stack;
+
+scheduler scheduler::s_instance(nullptr, kutil::allocator::invalid);
+
+const uint64_t rflags_noint = 0x002;
+const uint64_t rflags_int = 0x202;
+
+extern "C" {
+	void ramdisk_process_loader();
+	uintptr_t load_process_image(const void *image_start, size_t bytes, process *proc);
+};
+
+extern uint64_t idle_stack_end;
+
+scheduler::scheduler(lapic *apic, kutil::allocator &alloc) :
+	m_apic(apic),
+	m_next_pid(1),
+	m_process_allocator(alloc)
+{
+	auto *idle = m_process_allocator.pop();
+	uint8_t last_pri = num_priorities - 1;
+
+	// The kernel idle task, also the thread we're in now
+	idle->pid = 0;
+	idle->ppid = 0;
+	idle->priority = last_pri;
+	idle->rsp = 0;  // This will get set when we switch away
+	idle->rsp3 = 0; // Never used for the idle task
+	idle->rsp0 = reinterpret_cast<uintptr_t>(&idle_stack_end);
+	idle->pml4 = page_manager::get_pml4();
+	idle->quanta = process_quanta;
+	idle->flags =
+		process_flags::running |
+		process_flags::ready |
+		process_flags::const_pri;
+
+	m_runlists[last_pri].push_back(idle);
+	m_current = idle;
+
+	bsp_cpu_data.rsp0 = idle->rsp0;
+	bsp_cpu_data.tcb = idle;
+}
+
+uintptr_t
+load_process_image(const void *image_start, size_t bytes, process *proc)
+{
+	// We're now in the process space for this process, allocate memory for the
+	// process code and load it
+	page_manager *pager = page_manager::get();
+
+	log::debug(logs::task, "Loading task! ELF: %016lx [%d]", image_start, bytes);
+
+	// TODO: Handle bad images gracefully
+	elf::elf image(image_start, bytes);
+	kassert(image.valid(), "Invalid ELF passed to load_process_image");
+
+	const unsigned program_count = image.program_count();
+	for (unsigned i = 0; i < program_count; ++i) {
+		const elf::program_header *header = image.program(i);
+
+		if (header->type != elf::segment_type::load)
+			continue;
+
+		uintptr_t aligned = header->vaddr & ~(memory::frame_size - 1);
+		size_t size = (header->vaddr + header->mem_size) - aligned;
+		size_t pages = page_manager::page_count(size);
+
+		log::debug(logs::task, "  Loadable segment %02u: vaddr %016lx  size %016lx",
+			i, header->vaddr, header->mem_size);
+
+		log::debug(logs::task, "         - aligned to: vaddr %016lx  pages %d",
+			aligned, pages);
+
+		void *mapped = pager->map_pages(aligned, pages, true);
+		kassert(mapped, "Tried to map userspace pages and failed!");
+
+		kutil::memset(mapped, 0, pages * memory::frame_size);
+	}
+
+	const unsigned section_count = image.section_count();
+	for (unsigned i = 0; i < section_count; ++i) {
+		const elf::section_header *header = image.section(i);
+
+		if (header->type != elf::section_type::progbits ||
+			!bitfield_has(header->flags, elf::section_flags::alloc))
+			continue;
+
+		log::debug(logs::task, "  Loadable section %02u: vaddr %016lx  size %016lx",
+			i, header->addr, header->size);
+
+		void *dest = reinterpret_cast<void *>(header->addr);
+		const void *src = kutil::offset_pointer(image_start, header->offset);
+		kutil::memcpy(dest, src, header->size);
+	}
+
+	proc->flags &= ~process_flags::loading;
+
+	uintptr_t entrypoint = image.entrypoint();
+	log::debug(logs::task, "  Loaded! New process rip: %016lx", entrypoint);
+	return entrypoint;
+}
+
+process_node *
+scheduler::create_process(pid_t pid)
+{
+	kassert(pid <= 0, "Cannot specify a positive pid in create_process");
+
+	auto *proc = m_process_allocator.pop();
+	proc->pid = pid ? pid : m_next_pid++;
+	proc->priority = default_priority;
+	return proc;
+}
+
+void
+scheduler::load_process(const char *name, const void *data, size_t size)
+{
+	auto *proc = create_process();
+
+	uint16_t kcs = (1 << 3) | 0; // Kernel CS is GDT entry 1, ring 0
+	uint16_t cs = (5 << 3) | 3;  // User CS is GDT entry 5, ring 3
+
+	uint16_t kss = (2 << 3) | 0; // Kernel SS is GDT entry 2, ring 0
+	uint16_t ss = (4 << 3) | 3;  // User SS is GDT entry 4, ring 3
+
+	// Set up the page tables - this also allocates an initial user stack
+	proc->pml4 = page_manager::get()->create_process_map();
+
+	// Create an initial kernel stack space
+	void *sp0 = proc->setup_kernel_stack();
+	uintptr_t *stack = reinterpret_cast<uintptr_t *>(sp0) - 7;
+
+	// Pass args to ramdisk_process_loader on the stack
+	stack[0] = reinterpret_cast<uintptr_t>(data);
+	stack[1] = reinterpret_cast<uintptr_t>(size);
+	stack[2] = reinterpret_cast<uintptr_t>(proc);
+
+	proc->rsp = reinterpret_cast<uintptr_t>(stack);
+	proc->add_fake_task_return(
+			reinterpret_cast<uintptr_t>(ramdisk_process_loader));
+
+	// Arguments for iret - rip will be pushed on before these
+	stack[3] = cs;
+	stack[4] = rflags_int;
+	stack[5] = initial_stack;
+	stack[6] = ss;
+
+	proc->rsp3 = initial_stack;
+	proc->quanta = process_quanta;
+	proc->flags =
+		process_flags::running |
+		process_flags::ready |
+		process_flags::loading;
+
+	m_runlists[default_priority].push_back(proc);
+
+	log::debug(logs::task, "Creating process %s: pid %d  pri %d", name, proc->pid, proc->priority);
+	log::debug(logs::task, "     RSP  %016lx", proc->rsp);
+	log::debug(logs::task, "     RSP0 %016lx", proc->rsp0);
+	log::debug(logs::task, "     PML4 %016lx", proc->pml4);
+}
+
+void
+scheduler::create_kernel_task(pid_t pid, void (*task)())
+{
+	auto *proc = create_process(pid);
+
+	uint16_t kcs = (1 << 3) | 0; // Kernel CS is GDT entry 1, ring 0
+	uint16_t kss = (2 << 3) | 0; // Kernel SS is GDT entry 2, ring 0
+
+	// Create an initial kernel stack space
+	proc->setup_kernel_stack();
+	proc->add_fake_task_return(
+			reinterpret_cast<uintptr_t>(task));
+
+	proc->pml4 = page_manager::get()->get_kernel_pml4();
+	proc->quanta = process_quanta;
+	proc->flags =
+		process_flags::running |
+		process_flags::ready;
+
+	m_runlists[default_priority].push_back(proc);
+
+	log::debug(logs::task, "Creating kernel task: pid %d  pri %d", proc->pid, proc->priority);
+	log::debug(logs::task, "    RSP0 %016lx", proc->rsp0);
+	log::debug(logs::task, "     RSP %016lx", proc->rsp);
+	log::debug(logs::task, "    PML4 %016lx", proc->pml4);
+}
+
+void
+scheduler::start()
+{
+	log::info(logs::task, "Starting scheduler.");
+	wrmsr(msr::ia32_gs_base, reinterpret_cast<uintptr_t>(&bsp_cpu_data));
+	m_tick_count = m_apic->enable_timer(isr::isrTimer, quantum_micros, false);
+}
+
+void scheduler::prune(uint64_t now)
+{
+	// Find processes that aren't ready or aren't running and
+	// move them to the appropriate lists.
+	for (auto &pri_list : m_runlists) {
+		auto *proc = pri_list.front();
+		while (proc) {
+			bool running = proc->flags && process_flags::running;
+			bool ready = proc->flags && process_flags::ready;
+			if (running && ready) {
+				proc = proc->next();
+				continue;
+			}
+
+			auto *remove = proc;
+			proc = proc->next();
+			pri_list.remove(remove);
+
+			if (!(remove->flags && process_flags::running)) {
+				auto *parent = get_process_by_id(remove->ppid);
+				if (parent && parent->wake_on_child(remove)) {
+					m_blocked.remove(parent);
+					m_runlists[parent->priority].push_front(parent);
+					m_process_allocator.push(remove);
+				} else {
+					m_exited.push_back(remove);
+				}
+			} else {
+				m_blocked.push_back(remove);
+			}
+		}
+	}
+
+	// Find blocked processes that are ready (possibly after waking wating
+	// ones) and move them to the appropriate runlist.
+	auto *proc = m_blocked.front();
+	while (proc) {
+		bool ready = proc->flags && process_flags::ready;
+		ready |= proc->wake_on_time(now);
+
+		auto *remove = proc;
+		proc = proc->next();
+		if (!ready) continue;
+
+		m_blocked.remove(remove);
+		m_runlists[remove->priority].push_front(remove);
+	}
+}
+
+void
+scheduler::schedule()
+{
+	// TODO: lol a real clock
+	static uint64_t now = 0;
+
+	pid_t lastpid = m_current->pid;
+
+	m_runlists[m_current->priority].remove(m_current);
+	if (m_current->flags && process_flags::ready) {
+		m_runlists[m_current->priority].push_back(m_current);
+	} else {
+		m_blocked.push_back(m_current);
+	}
+
+	prune(++now);
+
+	uint8_t pri = 0;
+	while (m_runlists[pri].empty()) {
+		++pri;
+		kassert(pri < num_priorities, "All runlists are empty");
+	}
+
+	m_current = m_runlists[pri].pop_front();
+
+	if (lastpid != m_current->pid) {
+		task_switch(m_current);
+
+		bool loading = m_current->flags && process_flags::loading;
+		log::debug(logs::task, "Scheduler switched to process %d, priority %d%s.",
+				m_current->pid, m_current->priority, loading ? " (loading)" : "");
+	}
+}
+
+void
+scheduler::tick()
+{
+	if (--m_current->quanta == 0) {
+		m_current->quanta = process_quanta;
+		schedule();
+	}
+	m_apic->reset_timer(m_tick_count);
+}
+
+process_node *
+scheduler::get_process_by_id(uint32_t pid)
+{
+	// TODO: this needs to be a hash map
+	for (auto *proc : m_blocked) {
+		if (proc->pid == pid) return proc;
+	}
+
+	for (int i = 0; i < num_priorities; ++i) {
+		for (auto *proc : m_runlists[i]) {
+			if (proc->pid == pid) return proc;
+		}
+	}
+
+	for (auto *proc : m_exited) {
+		if (proc->pid == pid) return proc;
+	}
+
+	return nullptr;
+}